The epidemiology of P. falciparum and P. vivax in East Sepik Province, Papua New Guinea, pre- and post-implementation of national malaria control efforts

and investigate two cross-sectional surveys conducted in national and 2 and 2011-12). Differences between studies were investigated using chi-square (χ²), Fischer’s exact tests and Student’s t-test. Multivariable logistic regression models were built to investigate factors associated with infection at individual and household level.


Abstract Background
In the past decade, national malaria control efforts in Papua New Guinea (PNG) have received renewed support, facilitating nationwide distribution of free long-lasting insecticidal nets (LLINs), as well as improvements in access to parasite-con rmed diagnosis and effective artemisinin-combination therapy in 2011-12.

Methods
To study the effects of these intensi ed control efforts on the epidemiology and transmission of P. falciparum and P. vivax infections and investigate risk factors at the individual and household level, two cross-sectional surveys were conducted in the East Sepik Province of PNG; one in 2005, before the scaleup of national campaigns and one in late 2012-early 2013, after 2 rounds of LLIN distribution (2008 and 2011-12). Differences between studies were investigated using chi-square (χ²), Fischer's exact tests and Student's t-test. Multivariable logistic regression models were built to investigate factors associated with infection at the individual and household level.

Results
The prevalence of P. falciparum and P. vivax in surveyed communities decreased from 55% (2005) to 9% (2013) and 36% to 6%, respectively. The mean multiplicity of infection (MOI) decreased from 1.8 to 1.6 for P. falciparum (p=0.08) and from 2.2 to 1.4 for P. vivax (p<0.001). Alongside these reductions, a shift towards a more uniform distribution of infections and illness across age groups was observed but there was greater heterogeneity across the study area and within the study villages. Microscopy positive infections and clinical cases in the household were associated with high rate infection households (>50% of household members with Plasmodium infection).

Conclusion
After the scale-up of malaria control interventions in PNG between 2008 and 2012, there was a substantial reduction in P. falciparum and P. vivax infection rates in the studies villages in East Sepik Province. Understanding the extent of local heterogeneity in malaria transmission and the driving factors is critical to identify and implement targeted control strategies to ensure the ongoing success of malaria control in PNG and inform the development of tools required to achieve elimination. In household-based interventions, diagnostics with a sensitivity similar to (expert) microscopy could be used to identify and target high rate households.
Background past decade, renewed malaria control efforts initiated by the PNG National Department of Health with support from The Global Fund to ght AIDs, TB and malaria have once again signi cantly reduced the burden of malaria in PNG (18)(19)(20)(21)(22). This National Malaria Control Program (NMCP) has focused on the nationwide distribution of free LLINs on a 3-yearly cycle since 2008, greater availability of prompt diagnosis and effective treatment through the introduction of rapid diagnostic tests (RDTs) and artemisinin combination therapies (ACTs) in 2011-2012 and a behaviour change campaign (18, 21,23). ESP was among the rst Provinces in PNG where the LLIN distribution campaign was implemented, which together with the availability of historical data makes it an ideal location to investigate the changing epidemiology of malaria in PNG.
While the previously reported reduction in nationwide prevalence and incidence of malaria and all-cause mortality rates in young children are signi cant achievements (18, [21][22][23][24], important questions related to the nature of the epidemiological transition and long-term impact of intensi ed control on parasite-hostvector interactions remain unresolved. Sustained reduction in malaria transmission can lead to a decrease in naturally acquired immunity and consequently, a shift in the peak burden of malaria infection and illness to older age groups and change other risk factors. A declining burden of malaria illness and high-density parasite infections can also mask a substantial community burden of low-density but gametocytaemic infections that sustain transmission (25)(26)(27). In addition, the possible differential effects of these interventions on the various Plasmodium species is important to consider in co-endemic regions such as PNG, given the distinct biological differences of the two main Plasmodium species.
While declining prevalence and morbidity have been documented on a regional level, the effects of decreasing malaria transmission on a smaller geographical scale remain to be investigated. Heterogeneities in transmission and disease burden have been described at various scales in PNG (4,21,25) and can be a major challenge to further reducing the burden of malaria, as hotspots can act as the source of infection for other neighbouring areas (28), and especially P. vivax can be resilient (25).
To investigate the impact of the renewed malaria control efforts on the epidemiology of malaria within communities, data from two cross-sectional community surveys conducted in the Maprik and Wosera-Gawi districts of East Sepik Province, PNG; one in April -May 2005, before the implementation of national LLIN distribution campaign and one in October 2012 -February 2013, were directly compared. Risk factors for parasite prevalence, and spatial heterogeneity were investigated in addition to gametocytaemia and complexity/multiplicity of infection. Understanding the key factors related to heterogeneity and residual malaria transmission is critical to ensure sustained malaria control in PNG and particularly, when starting to consider sub-national elimination strategies.

Study design
Two cross-sectional community surveys were conducted in the Maprik and Wosera-Gawi districts in the East Sepik Province of PNG; one in April -May 2005 (9), before the implementation of national LLIN distribution campaign and one in October 2012 -February 2013. The 1st round of LLIN distribution in ESP took place in 2008, followed by a second round in 2011-2012. LLINs in the study area in ESP were distributed at most 12 months prior to the 2012/13 survey: September 2011 (Brukham, Ulupu and Ilahita catchement areas) and October 2012 (Wombisa and Burui catchement areas). Introduction of a test-andtreat approach and a switch to artemether-lumefantrine (AL) as rst-line treatment and formal adoption of 14 days 0.25 mg/kg primaquine for vivax-con rmed malaria occurred at the end of 2011/beginning of 2012. The Maprik and Wosera-Gawi districts, consist of an area of over 160 km 2 with low hills, plains and riverine plains with a wet tropical climate (3,9). The natural vegetation is lowland hill forest that has mostly been replaced by re-growth following cultivation and wide grasslands on the plains near the Sepik River. The people in this area live in villages of hundred to several hundred individuals and the villages are sometimes divided in geographically distinct hamlets. The majority of the people live from subsistence farming. There are several government health centres, church health centres and smaller aid posts in the area and the referral hospital is in Maprik.
All eligible and willing residents of households in the local communities in the study area were invited to participate and following written informed consent, demographic information (age, gender, familial relationship, bed net use), history of febrile illness, household or village location by GPS receiver and blood samples were collected. The design of both studies was similar, and details of the 2005 study have been previously described (9). All details of the recent cross-sectional are described below. In the 2005 study questions related to bednets were not directed to insecticide treated nets (ITNs) or LLINs in particular, and are likely to be untreated nets. In 2012-13, after the start of the large-scale LLIN campaigns, questions were asked speci cally about LLINs and not ITNs or untreated nets. Therefore, when variables are reported where the 2005 study is included, the term bednets will be used, while in the 2012-13 study the term LLINs can be applied.
Capillary blood (250-300µL) was collected into K + EDTA microtainers, thick/thin lms were prepared and haemoglobin levels were measured (Hemocue). The collected blood was centrifuged, the plasma removed and stored at -80ºC, and the red cell pellet stored at -20ºC until DNA extraction. If febrile illness was reported, a rapid malaria diagnostic test (Carestart Pf/Pan) was performed and those positive by RDT treated with Artemether-Lumefantrine (Coartem). For RNA preservation, 50µL of whole blood was immediately transferred to a tube containing 250µL of RNAProtect (Qiagen) and stored at -80ºC within 8 hours of collection until RNA extraction.

Plasmodium spp. detection
Giemsa-stained thick and thin lms from the 2005 survey were examined by LM (minimum of 200-high powered elds) by two independent experienced microscopists, with discrepancies adjudicated by a third independent microscopist, as previously described (9). All slides collected during the 2012/13 study, were similarly examined once by experienced microscopists. Slides from LM and/or PCR positive individuals and a random selection of negative slides were re-examined by an independent microscopist with discrepancies adjudicated by a third WHO-certi ed Level 1 or 2 microscopist. Parasite densities were calculated from the number of parasites per 200 or 500 white blood cells (WBCs) (depending on parasitaemia) and an assumed total peripheral WBC count of 8,000/µL (29), with the nal density taken as the geometric mean of the two values.
The presence of parasite DNA in all blood samples was investigated by MM. In 2005 DNA was extracted from red blood cell pellets using QIAmp 96 DNA Blood kits (Qiagen) and a post-PCR, ligase detection reaction/microsphere assay (LDR-FMA) was used to determine the presence of P. falciparum, P.vivax, P. malariae and P. ovale (9,30). In the 2012/13 survey, DNA was extracted from the equivalent of 200µL whole blood using the Favorgen 96-well Genomic DNA Extraction Kit (Favorgen) following the manufacturer's instructions and eluted in 200µL. Initially a generic quantitative PCR (QMAL) that ampli es a conserved region of the 18S rRNA gene was run on all samples (31). Species-speci c quantitative PCRs (qPCR) were then performed on all positive samples as described (32). The qPCR was previously directly compared to the LDR-FMA and assays are approximately equivalent in sensitivity and speci city (32). Copy numbers were quanti ed based on serial dilutions of plasmid controls run in parallel.

Gametocyte detection
The number of observed P. falciparum (2005 and 2012/13) and/or P. vivax (2012/13 only) gametocytes were recorded separately from asexual stages during microscopic examination of all blood slides.
Gametocyte detection by qRT-PCR was performed on samples from 2012/13: RNA was extracted from all P. falciparum and/or P. vivax qPCR-positive samples using the Qiagen RNeasy plus 96 kit, according to the manufacturer's procedures. A genomic DNA removal by gDNA eliminator columns and DNase (Qiagen) step was included in the procedure. Absence of gDNA was con rmed by qPCR and presence of parasite RNA after extraction was veri ed by qRT-PCR with the same primers and probe as the QMAL qPCR described above. P. falciparum and P. vivax gametocytes were detected by qRT-PCR of the highly expressed gametocyte markers pfs25 and pvs25 as previously described (31). The limit of detection of the pfs25 and pvs25 qRT-PCRs in the laboratory set up in PNG, were 6 copies/transcript per reaction for both assays.

Genotyping
To genotype P. falciparum and P. vivax LDR-FMA/qPCR positive samples, high-resolution Pfmsp2, Pvmsp1f3 and PvMS2 genotyping was performed as previously described (33,34). Brie y, for P. falciparum a nested multiplex PCR approach was used to amplify 3D7 and/or FC27 family alleles of Pfmsp2 using family-speci c primers labelled with a uorescent dye, or a multiplex primary PCR was used to amplify the markers Pvmsp1F3 and PvMS2 with uorescently labelled primers. The PCR products were analysed by 1.5% agarose gel electrophoresis and the number and size of alleles were then determined by capillary electrophoresis using a 23 ABI 3730XLs platform (Macrogen, Korea) with the internal size standard GSLIZ500 and data was subsequently analysed using GeneMarker 2.4.0 demo version (SoftGenetics). The multiplicity of infection (MOI) for P. falciparum was de ned as the number of Pfmsp2 alleles counted within a sample, and P. vivax MOI was de ned by the number of Pvmsp1f3 or PvMS2 alleles, whichever was higher.

Data analysis
Raw data from the 2005 survey (9) was reanalysed to compare prevalence and predictors with data from the 2012/13 survey. Clinical malaria cases were de ned as febrile illness (current or previous 48 hours) in the presence of P. vivax or P. falciparum asexual parasites by LM (any density). Symptomatic infections were de ned as febrile illness (current or previous 48 hours) in the presence of parasites as detected by LM and/or MM. Parasite densities were log transformed and the geometric mean per µL whole blood are reported. Four categories were used to describe anaemia: non-anaemia, mild, moderate or severe, which were de ned based on haemoglobin concentrations (measured by Hemocue) and strati ed by age and sex as per the WHO recommendations (Additional le 1) (35). On average, there were 3.7 people in a household (3.8 in 2005; 3.5 in 2012-13). A high-infection rate household was de ned on the basis of having more than 50% of household members with a Plasmodium infection (determined by molecular method) Differences between the two studies were investigated using chi-square (χ²) and Fischer's exact tests for categorical characteristics and Student's t-test for normally-distributed continuous variables. Tests were two-tailed and the con dence level was set at 95%. Univariable analysis of factors associated with P. falciparum or P. vivax infection (determined by MM and/or LM) and high rate household were conducted using logistic regression. Multivariable logistic regression models were built with variables with p < 0.15 in univariate analysis and variables were selected using likelihood methods for models with minimal Akaike information criterion (AIC). Analyses were performed using Stata 12. Genetic diversity analyses were done using FSTAT software version 2.9.3.2 to de ne allele frequencies, and the expected heterozygosity (He). Household, village and health facility location data was collected using a handheld GPS receiver (Garmin GPSmap62sc). Maps were constructed with collected GPS coordinates using ARC GIS Pro 10.4

Study characteristics
A total of 2,744 participants from 15 villages in ve distinct geographical areas of East Sepik Province participated in the 2005 cross-sectional survey. Of these 121 (4.4%) were excluded because of missing demographic or LM data, while insu cient nger-prick blood sample for LDR-FMA analysis led to exclusion of an additional 96 (3.5%) individuals. Overall 2,527 participants from 659 households with complete demographic and infection status data were available for comparison with 2012/13 data. A total of 2500 participants in 14 villages from the same areas participated in the 2012/13 cross-sectional survey. Of these, 14 (0.6%) were excluded from the analysis because of missing LM results. Overall, 2,486 participants from 704 households with complete demographic and infection status data were available.
The two studies were very comparable in their design, however, they differed in the demographic and clinical characteristics of the participants (Table 1 Table 1). The prevalence of clinical episodes of malaria (P. falciparum and/or P. vivax) had dropped from 3.6-2.3% (p = 0.007). There was a lower prevalence of anaemia in 2012/13, and accordingly, mean haemoglobin levels were signi cantly higher in 2012/13 (11.1 ± 1.8 g/dL) compared to 2005 (10.5 ± 1.7 g/dL; p < 0.001) ( Table 1). In addition, a higher proportion of individuals reported a current or recent febrile illness in 2012/13 (14.4% vs 7.4% p < 0.001).   Table 2). The presence of P. falciparum and P. vivax gametocytes in the 2012/13 survey was also investigated with a qRT-PCR detecting gametocyte speci c RNA, and the proportion of P. falciparum gametocyte positives by that method was signi cantly higher than with LM, 51.4% (p < 0,001, χ²). P. vivax gametocytes detected by LM were not recorded in 2005. In 2012/13 the prevalence of P. vivax gametocytes was 0.8% by LM and 0.9% by qRT-PCR (Table 2).

Comparing risk factors for Plasmodium spp. infections and illness
Village of residence was a risk factor for parasite infection in both surveys (Table 3). In addition, moderate or severe anaemia was a strong predictor for P. falciparum infection in both surveys, but not for P. vivax (Table 3). While males were more likely to be P. falciparum infected than women in 2005 (aOR 1.  (Table 3), and recent antimalarial treatment was associated with reduced risk of P. vivax in 2005. Infection with the other species was a predictor for both P. falciparum and P. vivax in 2012-13 (Table 3). Although there was a considerable reduction in the overall prevalence of malaria between 2005 and 2012/13, there was no obvious shift in the age of peak prevalence of P. falciparum and P. vivax infections (Fig. 1, Table 3). However, a more uniform distribution of infections across all age groups was observed in 2012/13 compared to 2005, with age no longer a signi cant predictor of P. falciparum infection in 2012/13 (Table 3) The decrease in the prevalence of P. falciparum and P. vivax was much greater in some villages than in others, indicative of heterogeneous transmission (Fig. 2).  was signi cantly associated with sub-microscopic P. vivax infections in both years (p < 0.001, logistic regression). Despite the signi cant decrease in prevalence, the overall genetic diversity (He) was high in both years for both species (Table 4). Despite large differences in prevalence in 2012-13, MOI and He were very similar across catchment areas (Table 4) and villages, and diversity seemed lowest in the highest prevalence area. In both years, the majority of P. falciparum infections in individuals over 12 years were single clone infections, whereas between the ages of 3 to 12 years more multiple clone infections were found (Fig. 3A). Single clone P. vivax infections were similarly distributed between all age groups in 2005, however, in 2012/13, there was a signi cantly higher proportion of P. vivax single clone infections in ages above 9 years (Fig. 3B). The proportion of multiple clone P. falciparum and P. vivax infections are quite evenly distributed in the different areas, despite differences in prevalence between the villages in both years ( Fig. 3C and 3D).

Heterogeneity of infection at the household level
Overall  Heterogeneity in malaria prevalence between villages is not a new phenomenon (3,4,(37)(38)(39). In the 2012-2013 study, transmission in some villages had not declined as much as in others, increasing this difference between and within villages. In those areas with higher prevalence, reported use of LLINs was lower and anaemia and fever were more prevalent. While coverage of bed nets was already high in this area in 2005 (88.3%), the majority of nets at that time were untreated nets, whereas in 2012-13 most people should have had access to LLINs, and questions speci ed LLINs. In the village with the highest density, Sunuhu, LLIN use was lowest, which could be a major factor contributing to the high level of transmission in that village. The use of insecticide-treated nets (ITNs) has been shown in many studies to be effective in reducing mortality and morbidity from malaria (40)(41)(42). In addition, it is thought that use of ITNs leads to community-level effects, where the majority of the population (even those not using ITNs) are protected when ITN coverage in the community is high, due to reduction in the number of infected mosquitoes and mosquito survival (43)(44)(45)(46). This effect has, however, not been well quanti ed and impact of LLINs varies with the coverage rate. In addition, the required coverage might be different for different areas, depending on local factors such as the anopheline density, species composition and both vector and human behaviour (47). In future studies, investigating the complex interplay of these different parasite, vector and human factors at the village level would facilitate a better understanding of the impact of bednets and other community-level interventions.
Renewed political and nancial will to strengthen malaria control at the beginning of the millennium, resulted in the PNG National Department of Health launching a new campaign to quickly achieve high levels of LLIN ownership and usage. Nationwide free LLIN distribution took place between 2004 and 2009 and resulted in a signi cant increase in ownership of bed nets (any type 80.1%; LLINs 64.6%) (20).
Despite this increase, reported LLIN use remained low (32.5%), and the majority of people not using nets reported not having access to (unoccupied) nets (20). A second round of country-wide LLIN distribution was conducted between 2010-2014 to cover the gaps in mosquito net coverage (21). LLINs in the study area in ESP were distributed 12 months prior to this 2012/13 survey (September 2011 and October 2012) and a new round of LLIN distribution occurred in 2014/2015 (personal communication, National malaria control program/Rotarians Against Malaria). Large-scale LLIN distribution campaigns were performed in ESP earlier than in many other provinces with a high malaria burden, and LLIN coverage in ESP seems to be higher than on the North Coast area in general (21,36), likely due to a higher density of nuisance biters encouraging greater use of LLINs. The substantial impact of LLINs on transmission in this area might also be due to the fact that transmission in ESP is predominantly driven by An. punctulatus, which feeds later at night (after midnight/early morning) and equally indoors and outdoors, making it highly susceptible to LLINs. A recent study reported an increase in prevalence of An. punctulatus and a decrease of An. koliensis, another late biter, in an adjacent area in East Sepik province, as well as a shift to earlier mean biting times of An. punctulatus in addition to signi cantly reduced man-biting-rates and annual entomological inoculation rates after the LLIN campaign (48). Proximity to vector-breeding sites is related to the risk of malaria and can also be a main driver of heterogeneity (49-51), but was not investigated in the current study Cultural factors, socio-economic status and education level play an additional role in risk of infection and can vary across villages and households, as well as human behavioural and genetic factors (9, 21, 28, 52-56). The time and manner of LLIN distribution in these areas can play a role in their availability, use and quality/age of the LLIN. For example, in some areas the LLINs might have been distributed directly to each household, whereas in other areas, people will have gone to their local health centres to obtain their LLINs. There are also real or perceived differences in the availability of RDTs, effective antimalarials and quality of care received at the health centres (19). In Sunuhu, for example, the population is of a different ethnic origin than surrounding villages, more closely related to the population in Gwanga local level government (LLG) than Ilahita LLG. The Sunuhu population have less material wealth, less access to nutritious food, and rates of malnutrition and generally poor health are more common than in neighbouring villages. They may also be less likely to access care at the nearby health centre (due to distance, cultural difference, perceived bene t etc.) and it's possible their access to LLINs has been reduced as a result.
Despite the decrease in prevalence and signi cant geographic heterogeneity, the genetic diversity of both P. falciparum and P. vivax appears to have been maintained at relatively high levels. Multiplicity of infection and especially the proportion of multiple clone infections has decreased. The proportion of multiple clone infections correlated well with prevalence and might be a suitable indicator of hotspot areas and areas of high transmission. Further investigation of markers that are not under selective pressure is required for a more detailed analysis of the impact of malaria control on genetic diversity, differentiation and population structure in this area (57,58).
Although the prevalence of malaria has decreased across all age groups and there is no marked shift in the peak age of infections (compared to the 2005 survey and others (5)), there is a relatively higher proportion of symptomatic infections in the 2012/13 survey). This, together with a 5-to 9-fold increase in the proportion of symptomatic infections in adults is an indication that there might be reduced or delayed acquisition of immunity. More detailed investigations on the effect of the decreased transmission intensity on the incidence (and complexity) of malaria infections and clinical malaria episodes, and age/exposure related acquisition of clinical immunity are being conducted in several longitudinal child cohorts in East Sepik and Madang Province.
The impact on the prevalence of P. vivax was similar or higher as on P. falciparum, in 10 of 12 villages surveyed in both years. Based on the biology of relapsing P. vivax infections from hypnozoites and the fact that neither LLINs nor ACTs act to prevent these relapsing infections it is generally thought that P. vivax burden is more resilient to these tools and that an equivalent impact may not be observed in the same timeframe as for P. falciparum. In the same area as this study, a series of cohort studies showed that while P. vivax clinical episodes declined at rates comparable to P. falciparum, force of blood stage infections and prevalence took longer to decline (59). The data presented here suggest that 8 years postscale-up of LLINs appears to be a su cient length of time for the hypnozoite burden to have been exhausted in most villages.
A limitation of the study was that the MM used to detect infections in the two time periods was not the same. However, a previous study directly comparing qPCR and LDR-FMA reported substantial agreement between the two methods (32). While the LDR-FMA detected slightly higher numbers of P. falciparum infections in that study (47% vs 41%) (32), this is not su cient to be responsible for the observed difference in prevalence between the two studies, which mirrors the drop in prevalence by LM. The decreased prevalence of sub-microscopic P. falciparum and P. vivax infections in the 2012/13 study is potentially in uenced by the difference between qPCR and LDR-FMA and thus the data from this earlier survey will appear to have a slightly higher proportion of sub-microscopic infections than when qPCR was used. In addition, 2nd reads of microscopy slides in 2012-13 were performed based on qPCR results, potentially resulting in a lower proportion of sub-microscopic infection.
This study was not a formal component of the monitoring and evaluation of the national malaria control program, with the primary aim to delve into the impact that reduced transmission is having on the epidemiology of malaria rather than assess the program itself. Prevalence reported in this study is much higher than provincial averages from the national reports (22,60), which to a large extent can be explained by the higher sensitivity of the MM that were used in these studies as compared to microscopy and RDTs. Molecular tools are much more sensitive at detecting low levels of parasitaemia and are therefore crucial to get a detailed insight on the prevalence of not only clinical disease, but also asymptomatic reservoirs of infection. Microscopic infections in the household, as well as gametocyte carriers were associated with high rate infection households in both years and in 2012-13 many high rate households contained clinical cases, highlighting the utility of clinical and microscopy-based surveillance to identify transmission foci that could be speci cally targeted with interventions aimed at reducing not only clinical cases, but the asymptomatic reservoir as well. Although the national program is very effective in determining the impact on a national level, this study facilitates the investigation at a different level of sensitivity, geographical scale and subsequent detection of ne-scale heterogeneity in transmission.
Identifying and targeting focal points and hotspots of malaria is highly relevant for malaria control, since these are likely to be the areas where residual malaria transmission will persist, and can become an obstacle in efforts to eliminate malaria (28, 61, 62). In addition they can play a catalysing role: hotspots fuel transmission within transmission foci, and interventions targeted at transmission hotspots therefore have the potential to reduce community-wide malaria transmission (28, 62). Many of the high burden villages found in 2012/13 were also areas of the highest prevalence in 2005, and it remains to be seen if these same areas are still high in prevalence in future. If these hot spots are consistently at the same location, implementation of control tools to target them will be much easier. In other regions in the world, it has been observed that hotspots are remarkably stable even when transmission intensity declines, although clinical incidence might vary with time (28, 63-65). In addition to scaling up conventional vector control tools such as LLINs and indoor residual spraying in these hotspot areas, other tools could be implemented that might prove more useful in these particular areas. Ongoing entomological studies in East Sepik and Madang provinces may advise on the suitability of additional vector control tools.
Alternatively targeting the parasite via interventions to reduce the infectious reservoir in these communities, such as mass screening and treatment (MSAT) and mass drug administration (MDA).
Modelling has shown that MDA targeting blood and liver stage drugs might be more effective in reducing P. falciparum and P. vivax prevalence than MSAT (66) and implementation of such strategies may be feasible to achieve in these relatively small communities.

Competing interests
The authors declare that they have no competing interests.