Asymptomatic and submicroscopic malaria infections in sugar cane and rice development areas of Ethiopia

Background Water resource development projects such as dams and irrigation schemes have a positive impact on food security and poverty reduction but might result in increased prevalence of malaria. Methods Two cross-sectional surveys were conducted in the dry and wet seasons in irrigated and non-irrigated clusters of Arjo sugarcane and Gambella rice development areas of Ethiopia in 2019. A total of 4464 and 2176 blood samples were collected from Arjo and Gambella. A subset of 2244 microscopy negative blood samples were analyzed by PCR. Results Prevalence by microscopy was 2.0% (88/4464) in Arjo and 6.1% (133/2176) in Gambella. In Gambella, prevalence was significantly higher in irrigated clusters (10.4% vs 3.6%) than in non-irrigated clusters (p < 0.001), but no difference was found in Arjo (2.0% vs 2.0%; p = 0.993). Level of education was an individual risk factors associated with infection in Arjo [AOR: 3.2; 95%CI (1.27–8.16)] and in Gambella [AOR: 1.7; 95%CI (1.06–2.82)]. While duration of stay in the area for < 6 months [AOR: 4.7; 95%CI (1.84–12.15)] and being a migrant worker [AOR: 4.7; 95%CI (3.01–7.17)] were risk factors in Gambella. Season [AOR: 15.9; 95%CI (6.01–42.04)], no ITN utilization [AOR: 22.3; 95%CI (7.74–64.34)] were risk factors in Arjo, and irrigation [AOR: 2.4; 95%CI (1.45–4.07)] and family size [AOR: 2.3; 95%CI (1.30–4.09)] risk factors in Gambella. Of the 1713 and 531 randomly selected smear negative samples from Arjo and Gambella and analyzed by PCR the presence of Plasmodium infection was 1.2% and 12.8%, respectively. P. falciparum, P. vivax, and P. ovale were identified by PCR in both sites. Conclusion Strengthening malaria surveillance and control in project development areas and proper health education for at-risk groups residing or working in such development corridors is needed.


Introduction
Water resource development projects such as dams and irrigation schemes have a positive impact on food security and poverty reduction (1).However, such projects also alter ecosystems and increase mosquito abundance.They could result in a change in the malaria transmission pattern from seasonal to perennial in particular in areas with unstable malaria transmission (2,3).The work opportunity in irrigation projects attracts a large number of migrant workers who might be at increased the risk of malaria.This is further exacerbated if migrant workers carry new strains such drug resistance parasites gene and also non-immune migrant works move to the area this may lead to malaria outbreak (4,5).
Different agro-ecosystems and crop production types have an impact on mosquito proliferation, and consequently, the intensity of malaria transmission.Depending on the number of crop cycles, irrigation-based farming may also extend the mosquito breeding season and lengthen the malaria transmission period (6).Major crop irrigation schemes in Africa include rice, sugarcane, cotton, wheat and vegetables.Among them, rice grown in ooded irrigation provides potential larval habitats (7).Although a sugarcane plantation needs irrigation for maximum growth, it does not need to be ooded condition (8).Furthermore, An. gambiae, the major malaria vector in Africa, prefers environments with direct sunlight for breeding, thus the dense vegetation cover by sugarcane development would make the environment undesirable (6).However, poorly maintained water canals used for sugarcane irrigation may produce hatching grounds for mosquitoes (9).Studies have shown inconsistent results of malaria prevalence following the implementation of irrigation schemes.Several studies found that when compared to non-irrigated villages, the prevalence of malaria was increased in irrigated or dam areas (10)(11)(12)(13)(14).Other studies have shown decreased malaria prevalence in irrigated areas (6, 15,16).Such discrepancies suggest that the effects of irrigation or dams on malaria transmission are poorly understood.
In malaria-endemic settings asymptomatic infections are abundant (17,18).The occurrence of asymptomatic infections poses serious implications for malaria control and elimination programs (19,20), as these individuals might carry gametocytes that contribute to the persistent transmission of malaria (21,22).Often, the majority of asymptomatic infections are submicroscopic and can only be detected using molecular methods (23).In endemic areas, malaria elimination will not be feasible with the existing interventions alone (22,24).Thus, additional strategies to detect asymptomatic carriers need to be considered.Therefore, the objective of the present study was to assess asymptomatic and submicroscopic malaria in sugarcane and rice development areas of Ethiopia

Study area
Data and blood samples were collected in sugarcane (Arjo) and rice (Gambella) development area of Ethiopia.The Arjo Didessa Sugarcane development irrigation is located in Jimma Arjo, Dabo Hanna and Bedele districts of East Wellega and Buno Bedele Zones of Oromia regional state.It is 540 km Southwest of Addis Ababa and is located at latitude 8.6 o N, longitude 36.4 o E, and; an altitude with a ranges of 1300 to 2280 m.The mean annual rainfall is 1477 millimeters, with bimodal rainfall (long rainy season June to September and short rainy season February to March) and with a relatively short dry season between December to January.
The Saudi Star Agricultural Development PLC is located in Abobo district of Gambella region in West Ethiopia.It is 811km west of Addis Ababa located at latitude 7.9 o N and longitude 34.5 o E, and; an altitude with a ranges of 400-500m.The annual rainfall in Abobo is ranges between 900 to 1200 mm, with unimodal rainfall (long rainy season May to October) and long dry and hot season occurs between November-April (25).
The total population of Abobo district was 26,080 living mainly on subsistence farming and shing.From thirteen Kebeles that surround the Saudi Star rice irrigation, six Kebeles were randomly selected for this study.Irrigated clusters were in Saudi Star Bravo (BRA), Saudi Star GRC (GRC) and Village-17 (V17).Non-irrigated clusters were in Terkudi (TER), Village-12 (V12) and Village-13 (V13) (Fig. 1).The rice irrigation is supplied by open gravity-fed earthen ditches from the main canal of Alwero dam.The non-irrigation clusters used rain-fed farming, cotton, maize, sorghum and shing.

Demographic and parasitological surveys
Demographic data and blood samples were collected simultaneously in eight irrigated and seven non-irrigated clusters of Arjo and in three irrigated and three non-irrigated clusters of Gambella rice irrigation (Fig. 1.).Irrigated clusters were within 1 kilometer (km) radius and non-irrigated clusters were above 1 km radius from the edge of irrigated areas, considering Anopheles mosquito ight range (26).The spatial coordinates and demographic data were collected using Open Data Kit (ODK).Surveys were conducted in March 2019 (dry season) and October 2019 (wet season), representing low and high malaria transmission season, respectively.Households were randomly selected from each cluster to maximize coverage so that the surveyed population would represent each clusters.All consenting individuals living in the selected households were included in the study.Socio-demographic characteristics such as age, sex, occupation, place of residence, level of education, data on preventive measures of indoor residual spraying (IRS), insecticide treated net (ITN) ownership, type and village GPS coordinate were collected in both surveys.As well, malaria symptoms such as fever and any malaria related symptoms (chills, sweating, headache, vomit, and abdominal pain) during or 48 hours prior to blood collection.Study participant with body temperature above 35 o c during or 48 hours prior to blood collection considered as febrile.
Capillary blood samples (300µL) were collected from all consenting participants for diagnosis by Rapid Diagnosis Tests (RDT), microscopy, and qPCR.For qPCR, dried blood spot (DBS) were prepared.The dried blood smear slides and the four dots of DBS samples were packed in slide box and plastic bag with desiccant, respectively.All samples were stored at room temperature then transported to Jimma University diagnostic and parasite culture laboratory for microscopic examination, DNA extraction and qPCR parasite determination.

Sampling and sample size
The sample size (n) calculation was done using the formula for estimating single proportion at 95% con dence interval (CI), (Z α/2 = 1.96).To attained maximum sample size 50% was assumed for prevalence (P).A total of 384 household (n) were generated using 5% marginal error (d).Then by considering 4.9 household size in Oromia and 3.8 in Gambella region (27) the nal number of study participants was calculated for each study site.Study households for cross-sectional parasitological survey (CPS) were identi ed through systematic sampling of every fth house to obtain an average of 26 to 30 households for Arjo and 60 to 64 households for Gambella from each clusters depending on the total population size of the clusters.

Blood lm
Thick and thin blood lms were prepared and stained with 10% Giemsa solution and examined under light microscope at Jimma University.Blood lms were considered negative if no parasites were detected in 200 elds of the thick blood lm.Thick blood lms were used for detection and quanti cation of the parasites.Thin blood lms were used for species identi cation.
Gametocytes and asexual stage parasites were counted against 200 and 500 white blood cells (WBCs), respectively, and densities (parasite/µl) was estimated using a factor of 8000 leukocytes/µl (28, 29).All positive slides and 10% of the negative slides were rechecked by a third technician.

DNA extraction
Chelex-100 resin was used for DNA extraction following an established protocol (30).One dot of the DBS was cut into piece of approximately 3-5mm using a puncher.The puncher was cleaned by punching paper ten times after every sample.The blotted lter paper was transferred to 1.5ml Eppendorf tube then 950µl phosphate buffered saline (PBS) and 50 µl 10% Saponin were added for the lysis of RBCs.After mixing the sample, it was incubated at 4 o c for 4 hours or overnight.The mixture was centrifuged at 14,000rpm for 10 minutes at room temperature and the supernatant was discarded.Any remnant of Saponin was removed by adding 1ml of PBS and centrifuge at 14,000rpm for about 5 minutes.The remaining PBS was discarded by centrifugation of the tube for about 15 seconds.The lter paper was left for 15 minutes to dry at room temperature.Subsequently, 150µl 20%Chelex suspension and 100µl distilled water was added to extract the parasite DNA by incubating the mixture at 95 o c in a water bath for 10 minute with vortexing the mixture every 2 minutes during the process of incubation.Finally, the mixture was centrifuged at 14,000rpm for 1 minute and the 200µl supernatant (DNA) transferred to 0.5ml tube and stored at -20 o C for PCR analysis.

Plasmodium species identi cation by qPCR
Genomic DNA of each sample was ampli ed using 18S rRNA genes based primers.Ampli cation was performed in a 12µl PCR reaction mixture containing 2 µl of genomic DNA, 6µl PerrfeCTa (2X), 0.4 µl forward and reverse primers of P. falciparum, P. vivax and P. ovale each (F-F, F-R, Pv-1, Pv-2, Po-1, and Po-2) then 0.5µl Pf-fam, 0.5µl Pv-vic, and 0.5 µl Po-Ned for P. falciparum, P. vivax and P. ovale TaqMan probe in 0.1 double distilled water (31,32) since there was P. ovale report in Northwest and Southeastern Ethiopia.Ampli cation reactions was performed in a 96 well Quant Studio 3, Applied Biosystems Real-Time PCR, with an initial denaturation at 50°C for 2 min and 95°C for 2 min, and 95°C for 3 sec, followed by 45 cycles at 60 o Cfor 30 sec.

Statistical analysis
Data was analyzed in JMP Pro (version 16 SAS Institute Inc., Cary, NC, USA).Univariate analysis was done on the associations between malaria prevalence and independent variables by regressing a single independent variable against Plasmodium positivity.Multivariate logistic regression run following stepwise backward selection of independent variables and removal of non-signi cant variables with p > 0.05.All signi cant variables remain in the nal model and the model selection was based on the Akaike's information criterion (AIC).The R statistical software package version (4.2.2) was used to compute W Mann−Whitney test, bar graph and box plot.Model assumption was tested for each models.

Result
Demographic information was obtained from 6640 individuals including 4464 (67.2%) from Arjo and 2176 (32.8%) from Gambella and can be found in Table 1.
There was signi cant difference in parasitemia load between symptomatic and asymptomatic Plasmodium infection by microscopy.

Univariate and multivariate analysis of risk factors for Plasmodium infection
In Arjo, sex, age group, duration of stay in the area, and migrant worker status were not signi cantly associated with infection.Individuals who were not o ce workers had signi cantly higher malaria infection rates than o ce workers [OR: 5.

Discussion
In Gambella, prevalence by microscopy was higher in irrigated vs. non-irrigated clusters, while no difference was observed in Arjo.In Arjo, prevalence was higher in the wet season, while this was the case only for non-irrigated sites in Gambella, and the effect was reversed in irrigated sites.Among the individual risk factors, educational level was signi cantly associated with Plasmodium infection in Arjo and Gambella.Whereas, duration of stay in the area, and being a migrant worker were signi cantly associated with Plasmodium infection in Gambella.Known household factors such LLINs utilization status, and high transmission season were signi cantly associated with Plasmodium infection in Arjo.In Gambella, irrigation and family size of the households were signi cantly associated with Plasmodium infection.
Asymptomatic sub-microscopic malaria is a major obstacle to malaria elimination due to its being a hidden reservoir and facilitate onward transmission.Therefore, malaria elimination requires targeting sub-microscopic carriers (23).Several studies showed a range of PCR prevalence in low transmission areas from 0 to 16.8 and 16.3 to 82.5 in high transmission areas respectively (18, [33][34][35][36].In our study, the prevalence of sub-microscopic Plasmodium infection was 1.2% in Arjo (low transmission setting) whereas, 12.8% in Gambella (high transmission setting).Studies in Ethiopia showed different prevalence of sub-microscopic malaria 2.4% (37), 3.3% (38), 9.7% (19),12.7%(39), 18.7% (40) and 19.2% (41).In our study (Arjo site), the low prevalence of sub-microscopic malaria infection might be due to the intensive use of primaquine (PQ) since 2018 targeting the elimination of malaria in low transmission settings.
Plasmodium ovale was reported in both study sites for the rst time.All infections were submicroscopic.Correct identi cation of parasite species and parasite distribution are important to facilitate proper diagnosis, treatment, case management and malaria elimination.Previously P. ovale was reported in Ethiopia in 1969, but it was not reported until 2013 and 2015 from northwestern (42) and southeastern Ethiopia (43).Two recent serological studies showed that there is evidence of the presence of P. ovale and P. malariae in different parts of the country (38, 44).
In Arjo, signi cant increment of malaria prevalence was not observed between sugarcane-irrigated and non-irrigated clusters however, rice irrigation showed 3-fold increase in malaria infection than non-irrigated clusters.Moreover, malaria infection doubled in the low transmission season (12.9%) than high transmission season (6.7%) in rice irrigation clusters.Similar ndings were also documented in two studies from east-central Tanzania (12,14).On the other hand, high malaria prevalence was recorded in sugarcane irrigated than rice irrigated in northern Tanzania (15).The discrepancy of the results might be due to the majority of study population were migrant workers in Tanzanian study.
Also, studies conducted in our study sites demonstrated that anopheline mosquito abundance was higher in rice-irrigated clusters than non-irrigated clusters and the sporozoite rate in rice-irrigated clusters was 10-fold higher than non-irrigated clusters in Gambella (45).Whereas, in Arjo, the anopheline mosquito density was 7 fold higher in sugarcane-irrigated clusters than in non-irrigated clusters during wet season (46).This could be the odor emanates from rice strongly attracts the female anopheline mosquitoes oviposition than the sugar cane (47,48) and the microhabitats formed by the ooded type of rice irrigation is favorable for mosquito proliferation during the low malaria transmission season.However, in Arjo sugarcane plantation, large proportion of irrigation system was sprinkler water for seedlings.Thus, the result might suggest that rice irrigation had higher contribution for persistent malaria transmission in the area.
In the present study, prevalence of infection was 16-fold higher in the wet season compared to dry season.This nding was complemented by an entomological study that showed higher mosquito abundance in wet season than the dry season in Arjo (46).In addition, a retrospective study which was conducted in Arjo showed higher malaria cases in wet season (49).Variation in the prevalence of malaria in terms of season in irrigated sites was observed in other studies (11,14,25,50,51).The difference might be due to the type of irrigation and type of crop cultivated.
Vector control is one of the most cost effective tool in reducing malaria transmission.This includes spraying IRS and ITN utilization.
Previously high coverage of IRS was done by the African IRS programme funded by USAID-PMI until 2010.At the moment, IRS is conducted by the national malarial elimination program of the Ethiopian government and the coverage was limited and targeted to malaria hotspot areas.However, IRS coverage was higher 37.6% in Arjo area compared to the 2015 regional coverage 29.8% (54).
Similarly, IRS coverage was double (42.4% ) in Gambella than 2015 regional coverage 21.5% (54).The mass blood survey was conducted seven months after IRS sprayed and the protective effect might be reduced due to dosage concentration, duration, and other factors could affect the protective e cacy of IRS.The protective effect of the IRS was not as effective as expected in this study.
Although, LLINs utilization was 32.2% in 2017 (53) and 40.3% in 2020 (54), poor utilization and no utilization of LLINs still signi cantly associated with Plasmodium infection in both irrigated and non-irrigated clusters in Arjo.The risk of Plasmodium infection was 10-fold increase in households who had a habit of poorly utilization of LLIN than households who used persistently.Additionally, households that did not utilize LLIN were 22 times at higher risk of Plasmodium infection than persistently utilizing in Arjo.This result is in line with the results of different studies (14,(55)(56)(57).In both Arjo and Gambella studies showed that those who never attended school were 3.2 and 1.7 times at higher risk of Plasmodium infection than those who had secondary and above education level respectively.Our nding is in consistent with ndings from other similar studies (58-60).Level of education had direct associated on the knowledge and utilization of malaria prevention tools.While preschool age children were 81% less likely to acquire Plasmodium infection than higher educational level [AOR: 0.19, 95%CI: (0.07-0.53) p = 0.0016] in Gambella study.Our ndings indicates that poor LLIN utilization could be correlated with the level of education.
Newly migrant population is more prone to malaria infection due to different reasons.In our study, being a migrant worker were 4.7 times more prone to malaria infection than the non-migrant residents.In addition, migrant workers who stayed in the area for < 6 months and 7 to 12 months were 4.7 and 5.6 times at risk of malaria infection than those who stayed for more than three years, respectively.Our nding is in line with ndings from several studies on migrant workers involved in commercial agricultural activities (5,55,(58)(59)(60)(61).In the era of malaria elimination human mobility has to be in account as a malaria risk group due to these groups might carry new strain, drug resistance parasite gene and if they travel from non-endemic area to endemic region this can cause an outbreak.One of the drawback of the 1950s and 1960s malaria eradication campaigns was unable to consider human mobility (62).Plasmodium infection in Gambella was also signi cantly associated with family size.Similar nding were observed in country wide study in Ethiopia (63) and India (64, 65).There is a possibility when the number of resident in a household increases, the olfactory cues to attract the anopheline mosquito become stronger and increases the chance to be bitten by the vector (66).

Conclusion
The co-existence of P. vivax, P. ovale and P. falciparum coupled with the high malaria prevalence recorded in irrigated areas indicated the signi cant impact of irrigation schemes on malaria transmission in the study areas.Poor ITN utilization being illiterate and population movement were the main malaria risk factors in the area.In addition, staying in irrigated less than 12 months and having a family size more than 5 were malaria risk factors.Therefore, strengthening malaria control and elimination; including proper utilization of vector control tools and health education for migrant population, should be implemented in water resource development projects such as irrigation schemes.

Declarations Figures
See image above for gure legend

Table 3 .
). Stepwise backward elimination of independent variables with the highest p-value resulted in the nal model, in which only level of education remained signi cant predictors of Plasmodium infection with full model AIC = 871.8and in the nal model AIC = 863.5 (Table4.).

Table 3
Univariate analysis of individual risk factors and malaria infection by microscopy, Ethiopia (March 2019 and October 2019)

Table 4
Multivariate analysis of individual risk factor and malaria infection by microscopy in Arjo and Gambella, Ethiopia (March 2019 and October 2019) In contrast, season and LLINs utilization were risk factors for malaria infection.Wet season [OR: 5.99, 95%CI (2.33, 15.41), p = 0.0001] and not utilizing LLIN [OR: 8.55, 95%CI (3.00, 24.34), p = 0.0001] were signi cantly associated with infection (Table5.).Stepwise backward elimination of independent variables with the highest p-value resulted in the nal model, in which season and LLIN utilization signi cant predictors of infection with full model AIC = 317 and in the nal model AIC = 300 (Table6.).
* :p < = 0.01,**: p < 0.001***, €: children less than 7 years old not yet start formal educationIn Arjo, household risk factors such as irrigation status, family size, roof material, oor material, wall materials, number of sleeping rooms in the house, number of LLINs in the house, and IRS were not associated with infection.

Table 5
Univariate analysis of household risk factor by microscopy, Ethiopia (March 2019 and October 2019) Multivariate logistic regression of household risk factors between irrigated and non-irrigated clusters by microscopy in Arjo and Gambella, Ethiopia (March 2019 and October 2019)

Table 3 .
). Sex was not signi cantly associated with Plasmodium infection.In the nal model, level of education, duration of stay in the area and migrant worker signi cant predictors of Plasmodium infection with full model AIC = 925.3and in the nal model AIC = 920.5 (Table4.)InGambella, household risk factors such as family size, roof material, number of sleeping rooms in the house, number of LLIN per household and a house that was treated with IRS in the preceding 12 months and season were not associated with Plasmodium infection in the univariate analysis.

Table 5 .
). Stepwise backward elimination of independent variables with the highest p-value resulted in the nal model with full model AIC = 514.7 and in the nal model AIC = 505.1.Households located in irrigated clusters and HH with a family size of > 5 were signi cant predictors of Plasmodium infection.Even if, number of LLIN per HH in the nal model it was not signi cantly associated to HH Plasmodium infection (Table6.).