High e cacy of microbial larvicides for malaria vectors control in the city of Yaounde Cameroon: a cluster randomised study

Christophe Antonio-Nkondjio (  antonio_nk@yahoo.fr ) Organisation de Coordination pour la lutte Contre les Endémies en Afrique Centrale (OCEAC) Patricia Doumbe-Belisse University of Yaoundé I Landre Djamouko-Djonkam University of Dschang Carmene Sandra Ngadjeu University of Yaoundé I Abdou Talipouo University of Yaoundé I Edmond Kopya University of Yaoundé I Roland Bamou University of Dschang Marie Paul Audrey Mayi University of Dschang Nadege Sonhafouo-Chiana University of Buea Diane Leslie Nkahe University of Yaoundé I Raymond Tabue National Malaria Control Programme, Ministry of Public Health Dorothy Achu Fosah National Malaria Control Programme, Ministry of Public Health Jude D Bigoga National Reference Unit for Vector Control, Nkolbisson-University of Yaounde I Parfait Awono-Ambene Organisation de Coordination pour la lutte Contre les Endémies en Afrique Centrale (OCEAC) Charles S. Wondji Department of Vector Biology Liverpool School of Tropical medicine Pembroke Place, Liverpool L3 5QA, UK


Introduction
Africa's population almost doubled during the last two decades, from about 665 million in 2000 to 1.1 billion in 2019 1 . This rapid demographic growth has resulted in a massive migration of the population from rural to urban areas. The rapid demographic changes in major sub-Saharan Africa cities which are also associated to large-scale unplanned urbanization including poor housing, poor drainage, inadequate waste management, multiplication of slums, have signi cantly in uenced the epidemiology of vector-borne diseases such as malaria and arboviruses [2][3][4] . Malaria remains an important public health problem across the world affecting both rural and urban areas [5][6][7][8] . According to the latest world malaria report, 229 million malaria cases were reported in 2019 9 . Twenty-nine countries account for 95% of malaria cases in the world and almost all are from sub-Saharan Africa 9 . Long-lasting insecticidal nets (LLINs) and indoor residual spraying (IRS) are considered as the cornerstone for malaria prevention 10 . The large-scale deployment of these tools permitted to avoid about 1.5 billion malaria cases and 7.6 million malaria deaths between 2000 and 2019 11 . Roughly 1.9 billion LLINs have been distributed in sub-Saharan Africa between 2004 and 2019 9 . It is estimated that about 68% of households in sub-Saharan Africa had at least one LLIN in 2019 this suggesting a terri c increase compared to 5% in 2000 9 . However, control efforts are still affected by the rapid expansion of insecticide resistance. Almost all sub-Saharan countries have reported resistance to all four of the most commonly used insecticide classes 12,13 . Resistance to pyrethroids the compound used for impregnating bed nets is widespread 12,14,15 . Based on insecticide resistance monitoring data, many countries are now adopting new strategies to manage insecticide resistance and improve malaria control. These include switching to new control tools such as the deployment of pyrethroid-piperonyl butoxide (PBO) nets 16 or the combination of different control tools or interventions 17, 18 .
Larval source management (LSM) has proven in the past to be highly effective for lowering malaria transmission and even eliminating malaria vectors and disease transmission [19][20][21] . Historical literature reveals that the use of anti-larval mosquito control measures contributed to all successful eradication efforts 22,23 . In Egypt, the use of larviciding in the 1940s resulted in the elimination of malaria and the vector An. arabiensis from the region of Assouan 23 . In the Zambia, the implementation of an integrated malaria control program relying primarily on anti-larval control measures, contributed to the reduction by 97% of the annual malaria incidence from 514/1000 to 16/1000 between 1929 and 1950 24 . Several studies reporting signi cant impact of larval control on malaria transmission or malaria morbidity have been registered across the continent 19,25,26 . However, despite these historical facts and new evidences on larviciding e cacy this intervention is still not largely implemented for malaria control in sub-Saharan Africa due to the limited number of unbiased studies on its e cacy or effectiveness 26,27 . The World Health Organization (WHO) issued an interim position on larviciding recommending its use in moderate to low transmission settings as a supplement to core interventions (LLINs and IRS) in areas where breeding habitats are few xed and ndable 10 . The intervention could be particularly indicated in urban settings or in highland areas where breeding habitats are less important and malaria transmission moderate. According to the WHO guidelines 28 , LSM could be integrated into malaria control or general mosquito abatement programmes once transmission has been reduced to low or moderate levels after the use of LLINs or IRS, or once these interventions have reached their maximum practical impact.
In Cameroon malaria remains an important public health problem. Between 2015 and 2018, the incidence of malaria cases increased across the country highlighting the need to intensify malaria control efforts 29,30 .
Treated nets are the main measure implemented by the government to prevent malaria attacks. So far there have been four distribution campaigns of treated nets to the population. It is estimated that about 80% of households own a bed net and that close to 60% use treated nets regularly 30 . Apart from LLINs which were introduced in the country in the 1990s, there have been two pilot larval control trials initiated in the country to control Culex quinquefasciatus populations. The rst one conducted in the 1990s in the city of Maroua which consisted of two treatments per year of all breeding habitats with Bacillus sphaericus as larvicide had a limited impact on the biting densities of Culex quinquefasciatus mosquito populations 31 . The second pilot study conducted in Yaounde registered a 64% reduction of Culex quinquefasciatus biting densities. However, because the authors did not included any control cluster the interpretation of their ndings was limited 32 .
The city of Yaounde has a landscape with an alternation of both highland and lowland areas with over 90% of breeding sites located in lowland settings and could be an excellent environment to practice larviciding 33,34 . The population is approximately 3 million inhabitants and is characterised by a low malaria transmission pattern 35,36 . There have been so far not enough attempts to control malaria vectors using interventions suited to the landscape and ecological situation of the environment. Generating evidences on the e cacy of larviciding in different epidemiological context could improve malaria control across Africa. In the course of the present study, a cluster randomised trial including 26 clusters of 2 to 4 km 2 each divided into 2 groups 13 in the intervention area and 13 in the non-intervention area was conducted to assess the impact of larviciding on malaria transmission in the city of Yaoundé. The study showed a reduction of 68% of adult anopheline densities and of 79% of the entomological inoculation rate.

Household characteristics
Baseline community and entomological surveys were conducted from February 2017 to July 2018. Some data deriving from these studies have been published previously 37-40 41-44 . Household characteristics were almost similar across the two study groups. Modern houses built up with cement (50% and 62.77%) and traditional houses constructed with mud, plank, and mix material (50% and 37.23%) were recorded. Most households (> 84%) owned at least a LLIN, 47% and 48% of households in the control and intervention area respectively had one LLIN for two people ( Table 1). The majority of households had an average of 6 to 10 persons per household. Close to 20% of houses had screens on windows. The number of houses with ceilings was also similar between the two groups. 2020. This pattern in uenced breeding habitats availability and distribution in the city ( Table 2).
The proportion of habitats found with early or late instar anopheles larvae at baseline was 13.32% (1150/8633) in intervention area and 18.66% (1551/8313) in non-intervention area. During the intervention period, only 0.80% of sites (1102/137120) were found with anopheline larvae after larviciding treatments whereas, in nonintervention areas, 7.52% of sites (1934/25729) were found with anopheline larvae. Taking into account the clustering by treatment group and by period, it appeared that larviciding treatment signi cantly reduced the chances of water bodies being colonised by anopheline larvae (OR= 0.15 95% CI = 0.07 -0.32; P < 0.0001). The number of breeding habitats with late instar anopheline larvae was also reduced by over 73%. When was considered the effect of larviciding treatments on culicine larvae, a signi cant reduction of breeding habitats with culicine larvae could also be noticed (OR = 0.37 95% CI = 0.32 -0.42; P < 0.0001). High uctuation in the monthly distribution of breeding habitats with anopheline larvae closely associated with the rainfall pattern was recorded ( Figure 1). 69.24% * Percent reduction = 100 -(Non LCI at baseline/LCI at baseline x LCI during intervention/non-LCI during intervention) x 100 (Non-larviciding intervention area (Non LCI), Larvicidng Intervention area (LCI)) Evolution of the frequency of breeding habitats with Anopheles larvae In addition to crude data analysis, a mixed linear modelling approach was used to better assess the impact of larviciding treatments. A total of 1131 measurements in both the intervention and non-intervention areas were taken into consideration for the modelling analysis. Results con rmed a signi cant reduction (P < 0.001) of breeding habitats with anopheline larvae in the intervention area compared to the non-intervention area, though a signi cant decline in the proportion of breeding habitats with Anopheles larvae for both non-intervention and intervention areas was generally observed with time (Table 3). Different factors including season, ooding, agricultural activities were found associated with a signi cant impact on the OR of the model (P < .005). In uence of physicochemical factors on anopheline larvae presence in breeding sites before and during intervention The possible in uence of some physicochemical factors on anopheline larvae distribution was checked at both baselines and during intervention to assess any effect of these factors on the density of anopheline in breeding sites before and during intervention. At the baseline, few parameters including sulphate (R 2 = +0.07 vs R 2 = -0.27), H 2 O 2 (R 2 = -0.09 vs R 2 = +0.28) and nitrate (R 2 = -0.48 vs R 2 = +0.07) showed different correlation pattern with anopheline larvae density in breeding sites in non-intervention vs intervention areas (Table 4). During the intervention several compounds including TDS (R 2 = -0.33 vs R 2 = +0.22), organophosphates (R 2 = -0.33 vs R 2 = +0.18) and sulphate (R 2 = -0.51 vs R 2 = +0.02) were found to express different and high correlation pattern with anopheline larvae density in breeding sites in non-intervention vs intervention areas. Although differences recorded were not signi cant, these factors could be confounding factors and need further assessment. Adult mosquito abundance A total of 6,664 anophelines were collected in the course of the study. Species collected included An. gambiae s.l., An. funestus and An. ziemanni (Table 5). A subsample of 2762 An. gambiae s.l., was processed by PCR and both An. coluzzii (88.42%) and An. gambiae (11.58%) were recorded. Within the An. funestus group, out of 299 mosquitoes processed, 280 (93.65%) were An. funestus s.s., and 19 (6.35%) were An. leesoni. In almost all districts An. coluzzii was the predominant species; followed by An. gambiae. No signi cant variation in the composition of An. gambiae and An. coluzzii before and during the intervention was recorded in both the intervention and non-intervention areas (P>0.20) (Figure 2). An. funestus was recorded in few sites and was particularly abundant in the site of Mendong located close to the periphery with large swamps. Adult vector density was higher at baseline than for subsequent years throughout intervention in both the intervention and non-intervention areas ( Figure 3). After launching the intervention, a steady decrease in vector density was recorded in the intervention area. The average density of anopheline collected in the nonintervention clusters varied from 0.42 anopheline/trap/night at baseline to 0.23 anopheline/trap/night during the intervention. In the intervention clusters the average density of anopheline collected by CDC light traps varied from 0.47 anopheline/trap/night at baseline to 0.082 anopheline/trap/night during intervention. Larviciding was associated with 68% reduction of adult anopheline biting density. The density of mosquitoes collected indoor and outdoor in control and intervention area also varied signi cantly. The highest reduction was recorded with mosquitoes biting indoor 68.27% vs 57.74% outdoor (Table 6). When was compared the impact of the intervention on An. gambiae s.l., and An. funestus the two main vectors species in Yaounde, it appeared that An. gambiae s.l., biting density was reduced by over 71% whereas An. funestus density was reduced by 40% (Table 7).   Modeling the effect of larviciding intervention using general estimating equations after adjusting for clusters, baseline data and months it appeared that at baseline, vector density and EIR in intervention and nonintervention area were readily comparable while the implementation of larviciding treatment signi cantly reduced the risk of being bitten by anopheline (P<0.05), all this during the entire intervention period (beginning (2018), midway (2019) and end of the study (2020)) ( Table 8). In uence of house characteristics on mosquito distribution in the intervention and non-intervention areas at baseline and during the intervention In the course of the study, various parameters enabling or preventing mosquitoes of getting into houses were assessed to check how they were affected by larviciding intervention. At baseline apart of the type of house (RR = 1.22 ± 0.19; P = 0.006) and the presence of holes in the wall (RR = 0.77 ± 0.14; P = 0.003), and the absence of breeding sites near houses (RR = 0.78 ± 0.2 ; P=0.04) the risk of being bitten by mosquitoes was similar between houses in the intervention and the non-intervention areas. During the intervention, almost all parameters measured were associated with a signi cant risk of being bitten by mosquitoes in the nonintervention compared to the larviciding intervention area ( Table 9). The risk of being exposed to mosquito bites was always twice higher in the non-intervention compare to the intervention (RR (range) = 1.58 -2.98; P< .001).
Parameters of the house (holes on the wall, absence of ceiling, absence of screens on windows) known to increase exposure to mosquito bites were found as contributing to a less important exposure risk in intervention area compared to non-intervention area.  (Table 10).

Discussion
This study's main objective was to assess the impact of larviciding on biting anopheline densities and malaria transmission intensity in the city of Yaounde. The present study used entomological outcomes as primary endpoint rather than epidemiological outcomes because of limited nancial means. In Yaounde, over 90% of households own at least a net and over 70% of the population report using net regularly 38  During the study, continual application of larvicide was conducted rather than seasonal (during the rainy season) as done elsewhere 25 . This regular application of the larvicide led to a high reduction of breeding habitats with anopheline larvae, the density of anopheline larvae and late instar stages. These gures are consistent with previous ndings 47, 48 . Although studies conducted so far in Yaoundé suggested seasonal malaria transmission pattern 35,37,39 , it is possible that transmission could be occurring at an undetectable rate in some period of the year due to the permanent presence of An. gambiae sl in the city and gametocyte carriers. This observation supports regular application of larvicide all year long at least during the rst years of the intervention. Analysis of the landscape of the city of Yaounde and transmission risk pattern also indicated a heterogeneous malaria risk with some districts more affected than others 37,40 and is in favor of emphasizing larviciding interventions. An. funestus was less intensely affected by the intervention compared to An. gambiae sl and could derive from the fact that An. funestus breed in water bodies covered by emerging vegetation which could reduce the quantity of larvicide granules getting to water surface and available for larvae whereas, An. gambiae s.l. is mainly found in water bodies without vegetation 49 . Limited impact of larviciding due to vegetation cover was reported in previous studies 50 .
Several physico-chemical parameters were monitored in the course of the study to assess their in uence on mosquito distribution or larviciding treatments e cacy. Some of them including organophosphate, sulphate, conductivity and TDS were found to display different correlation patterns with larval density in intervention compared to non-intervention areas and could translate possible interaction with the larvicide. The possible in uence of physico-chemical parameters on microbial larvicide e cacy deserves further assessment.
The composition of the anopheline fauna (particularly An. gambiae and An. coluzzii) did not change signi cantly in the intervention and non-intervention areas before and during the intervention, which could suggest similar susceptibility status to larvicide of the two species as earlier suggested for insecticides 51  A moderate reduction of adult Culex species biting density was recorded. The limited impact of larviciding treatments on this species could be due to the fact that these mosquitoes breed in different types of habitats such as pit latrines, which were not targeted during larviciding treatments. It may also be possible that the impact of larviciding treatments in drains which are also preferential breeding habitats for Culex could have been limited due to the presence of solid wastes and many hiding places which could have limited the distribution of larvicide in the water 40 . Culex mosquitoes in Yaounde have also been reported to display a high resistance pro le 41,57 .
As for houses, various factors allow mosquitoes to easily get in, including holes in walls, presence of opened eaves or absence of ceiling, which were proven to have a limited in uence on indoor biting mosquito's density during intervention, compared to the baseline period in intervention areas. Also, factors preventing mosquitoes from entering houses, such as presence of screens on windows or use of LLINs were found to induce better protection in areas where larviciding intervention was implemented compared to non-intervention areas. Better housing has always been regarded as a factor that could improve protection against mosquito bites in urban settings 43,[58][59][60] .
The impact of the use of the microbial larvicide VectoMax on non-target organisms was also monitored and no signi cant impact on the non-target microfauna (Cladocerans, Rotifers, Ostracods and Copepods) was recorded. A high diversity of the microfauna was instead recorded in intervention areas. The larvicide may be integrated in the food chain of some of these microorganisms. Since the study was limited to its effect on microfauna further studies are needed to assess the effect of this larvicide on aquatic macrofauna.
This study had some limitations. (i) Due to limited nancial resources, the study mainly focused on entomological outcomes as primary endpoints rather than epidemiological outcomes as generally done.
However, it provided a proof of concept that larviciding could be a suitable measure for reducing malaria transmission intensity in Yaounde. (ii) The study did not assess the impact of the intervention on epidemiological outcomes. As such, further studies should urgently assess the impact of larviciding on epidemiological outcomes such as malaria incidence and parasite prevalence for evidence-based decision making. (iii) The study relied on self-report assessment to measure LLIN coverage and use. This could have biased the interpretation of the added effect of larviciding on LLINs. (iv) The study did not assess the costeffectiveness of larviciding which is very important for policymakers.

Conclusion
This study sets out to advocate the fact that the use of larviciding as a complement to LLINs could be a viable solution for controlling malaria transmission in Yaounde, in a context of rapid expansion of insecticide resistance and outdoor malaria transmission. The study provided strong evidence supporting the use of larviciding as a main intervention in urban settings. Results obtained should be considered by national control programmes and local Government to implement tailored control approach to improve the ght against vectorborne diseases in urban settings. Further studies should be carried out to assess the impact of larviciding on epidemiological outcomes in Yaounde, the cost-effectiveness of larviciding with microbial larvicide and ways to involve community in vector control activities to ensure the sustainability of such interventions.

Study area
The study was conducted in Yaoundé the capital of Cameroon (3° 52' 12 N; 11° 31' 12 E). Yaounde is located 726 meters above sea level and receives up to 1700 mm of rainfall annually. It displays an equatorial climate with two rainy seasons extending from March to June and from September to November lasting 7 to 8 months.
Despite its geographical location in the equatorial forest domain, the extension of settlements has signi cantly reduced the forest cover mainly found in nearby rural areas. The city extends 20 km wide and about 25 km long. Yaounde landscape comprises highlands and lowlands areas crossed by several rivers. Lowland areas are exploited during the dry season for agriculture. Houses are built on both hill slopes and in lowlands. Main rivers crossing the city include rivers Mfoundi, Ekozoa, Biyeme and Mefou.

Study design and larviciding activities
The primary objective of the trial was to assess the effect of larviciding on anopheline mosquito densities and malaria transmission rate in Yaounde. A cluster randomised trial was conducted in twenty-six districts referred to as clusters. Thirteen clusters served as control whereas the thirteen remaining were the intervention areas.
Each cluster was an area of 2 to 4 km 2 crossed by a river system encompassing both lowland and highland areas. The lowland part for the majority of clusters was sparse and exploited for agriculture or with human constructions. The evaluation zone was situated at the center of each cluster always in the lowland area.
Clusters were separated from one another by a distance of 500m to 1 km to minimize mosquito spillover from non-intervention to interventionsites. Baseline entomological data were collected from all clusters for 17 months, from March 2017 to July 2018. After this period microbial larvicide was applied in 13 clusters for 27 months (September 2018 to November 2020) (Figure 4). Adult biting densities collected using CDC light traps were used as the primary outcome. At the baseline, it was noticed that >90% of households owned at least one LLIN, but only 58.5% had one LLIN for two people as requested by the WHO 30 . At the end of the baseline sampling period, all clusters were ranked according to adult anopheline biting density. Clusters with similar biting density were grouped into pairs and from each pair, one cluster was randomly selected as the intervention site and the other as control using a computer-assisted programme.
In intervention clusters, all water collection points were treated. It was assumed that when larvicide was applied to the entire cluster, the buffer zone and the fact that the evaluation was conducted at the centre of the cluster, could reduce mosquito spillover from non-intervention sites to intervention areas. Treatments were conducted Teams of three to four male adult applicators conducted the application of larvicide across each cluster.

Endpoints
To assess the impact of the LSM intervention we used as primary outcome adult anopheline biting density collected using CDC light traps. Secondary outcomes included the entomological inoculation rate, the infection rate, the presence of anopheline larvae in breeding habitats and larval density.

Larval vector abundance
During the study, all breeding habitats were identi ed and characterised. Their size, physico-chemical characteristics and the presence or absence of anopheline and culicine larvae were recorded every month.
During the intervention, water collection points were checked every week in the intervention area to nd out the number of habitats containing early and late instar larvae, to determine the effectiveness of larvicide application. Surveillance of treated breeding sites was conducted 48 hours after the treatment by a team of two people (different from those who undertook the treatment) who visited at least 50% of the treated area and all breeding habitats found with larvae were retreated. Checking larvae in breeding sites was also conducted in non-intervention sites once every month to capture the progression of mosquitoes in these sites. All water bodies encountered were geo-located using a Garmin eTrex® GPS and recorded in a GIS database for analysis.
Water bodies were analysed to check the presence of mosquito larvae. The immature stages of mosquitoes were collected using standard dipping technique 62 . Using a 350 ml deeper, three to ve dips were performed for small breeding sites of less than 1 m 2 ; and ve to ten dips for breeding sites of more than 1 m 2 . For some habitats such as tyres or footprints which could be too shallow during certain periods, larval collection was conducted using a pipette. The average larval density (N) was estimated by calculating the ratio of the number of larvae collected per dip (using a dipper with a volume of 350 ml). Once collected larvae were classi ed according to their stages: early instars larvae (L1 and L2), late instars (L3, L4) and pupae. Anopheline larvae were separated from the culicines using morphologically identi cation keys 63,64  Research and import permit for the use of VectoMax®G in Cameroon was granted by the Minister of Trade (Reference IF014167; IF021096; IF031126).

Data analysis
Data were collected on forms, checked rst to ensure they were lled comprehensively, then recorded in excel databases. Linear mixed models with random intercepts and Generalized Estimating Equations were used to assess the effect of larviciding treatment on the presence of anopheline larvae (early and late instars) in water collection points as well as adult anopheline density, infection rate and Entomological Inoculation Rate (EIR) respectively, adjusting for baseline data. In a preliminary analysis, follow-up curves for the non-intervention and intervention areas were constructed to visualize differences in the responses between the two sites. Average trends and local polynomial regressions of the presence of anopheline larvae (early and late instars) in water collections, anopheline density, infection rate and EIR with date of evaluation were also constructed separately for the different groups to further visualize these differences. We also estimated a null model with random intercept and calculated the intraclass correlation coe cient (ICC) associated with the presence of anopheline larvae, anopheline density, infection rate and EIR respectively. Generalized Estimating Equations were further used to describe the variation in Anopheles density, infection rate and EIR in the population under study, while controlling for baseline survey and groups. In all these cases, the identity link function with a Gaussian distribution was used, and we resorted to model with independent correlation structures since models in which within-cluster associations or correlations among the repeated measures were taken into account by de ning more complex "working" correlation structures (like the autoregressive or unstructured correlation) did not converge. All analyses were carried out with the R 4.0.2 software using the R packages nlme, ggplot2, plyr, lattice, car, effects, emmeans and data.table. Odds ratios and risk ratio were calculated and adjusted for the year of intervention, cluster and season. Binary logistic regression was used to assess the distribution between species and physicochemical parameters in intervention and non-intervention areas. The Entomological  Activity schedule graph for larviciding trial in Yaounde