Reducing Filth Fly Populations and Transmission of Diarrhoeal Pathogens in Harsh Settings: a Cross-over Trial in South Sudan
Reducing Filth Fly Populations and Transmission of Diarrhoeal Pathogens in Harsh Settings: a Cross-over Trial in South Sudan
Objectives The main trial objective is to evaluate the efficacy of larvicide (LARV) or baited fly traps (TRAP), or both interventions combined, on filth fly density and hence transmission of diarrhoeal diseases.
Primary Objective
• To evaluate the efficacy of LARV or TRAP, or both interventions combined, in reducing the number of cases of all-cause diarrhoea (primary epidemiological endpoint).
Secondary Objectives
Tertiary Objectives
Trial Design The proposed trial will be a 3-period, 3-group, interventional non-randomized cross-over trial with parallel control (Table 1). The unit of intervention will be an individual refugee camp. Each implementation phase will be 6 months, followed by a 1-month wash-out period.
The trial arms will be:
Study Setting This study will take place in Maban County, Upper Nile State, South Sudan, which is home to 4 refugee camps. Many refugees originate from Sudan and others from Ethiopia, Eritrea, and Somalia. The 4 camps host 218,439 refugees (104,784 in Doro, 57,605 in Yusuf Batil, 34,914 in Kaya and 21,136 in Gendrassa), according to the UN High Commissioner for Refugees (UNHCR) population estimates from 30 November 2025.
The refugee camps shelters are constructed from a mix of locally sourced materials. Most are temporary shelters made from tarpaulin and plastic sheeting, but some use a mixture of wood, mud, thatch and corrugated iron. The latrines primarily comprise polypropylene sheeting suspended by wooden sticks, with an open roof. They are rudimentary and built to be easily replaceable in case of flooding. Camps are designed with a set ratio of latrines to family shelters.
According to recent District Health Information Systems data, the major causes of death among refugee children arriving to the camps are measles and malnutrition. Malaria remains the leading cause of morbidity and mortality, accounting for 66% of outpatient consultations, 30% of admissions and 50% of deaths. Diarrhoeal diseases affect all refugee age groups but are worse among those under 5s, attributable to 10% of deaths in this age group.
Age Eligibility Criteria: See "Eligibility"
Arms and Interventions: See "Arms and Interventions"
ITNs are the current standard of care in the refugee camps in Maban County. These are distributed every 2-3 years to protect the population from malaria. ITN coverage is expected to be equitable between intervention and control camps. ITNs will not impact filth flies as they are not hematophagous or nocturnal.
The control arm will not receive a placebo intervention; therefore this will be a non-blinded trial at the implementation level.
Intervention arm 1 (LARV): Dimilin® GR-2 larvicide is an effervescent granule containing 2% diflubenzuron (20g/kg), an insect growth regulator (IGR). Diflubenzuron interferes with chitin synthesis resulting in the inhibition of mosquito (Anopheles, Aedes and Culex spp.) or fly (Musca domestica and Stomoxys calcitrans) ecdysis. This intervention will interrupt filth fly development at the larval / pupal stage, to reduce overall population density. All eligible pit latrines (with at least 3 standing walls) assigned to this intervention arm will be treated with Dimilin® GR-2 at 0.5 kg/10m2; granules will be sprinkled onto damp sewage per pit latrine. If waste is desiccated, this intervention will be diluted in water and applied as a spray. This intervention will be re-applied to pit latrines and open defecation sites, every month during each 6-month implementation phase.
Intervention 2 (TRAP): iCatchi® ProMax fly traps. iCatchi® ProMax fly traps are designed for commercial and residential use. These traps are 10 L containers which hold 3 L of proprietary fly bait, formulated to maximize fly attractiveness to ensure high capture rates. The container design minimizes spillage and allows for easy set up and maintenance in different environments, including agricultural settings, waste management sites and urban areas. Each TRAP requires re-baiting every 8 weeks. TRAP will be placed on the ground outdoors and flies enter the large black container through rectangular holes between the bait bowl and larger trap chamber and are unable to exit; they will eventually die from starvation / dehydration. During each implementation phase, in camps assigned to this intervention arm, one iCatchi® ProMax fly trap will be placed outside (<5 m) pit latrines at a set trap to pit latrine ratio and will be re-baited every 2 months.
To prevent filth fly contamination between camps, Doro and Kaya camps will receive the study interventions since they are more than 7 km apart (29.4 km), based on maximum filth fly dispersal range. Yusuf Batil and Gendrassa (2.6 km apart) will be the parallel control camps throughout the study period; Kaya is 11.6 km away from Yusuf Batil and 13 km from Gendrassa; Doro is 23.4 km from Gendrassa and 25.8 km from Yusuf Batil.
During each implementation phase, project staff will go pit latrine-to-pit latrine to supervise intervention deployment. At the beginning of each wash-out phase, project staff will go pit latrine-to-pit latrine to supervise removal of TRAP; the residual efficacy of LARV is 1 month, therefore this intervention will dissipate during the wash-out period.
Explanation for the Choice of Comparators:
According to the WHO Vector Control Advisory Group (WHO-VCAG), studies should always have a control group from which data collection occurs contemporaneously. Our trial design includes an untreated control arm, which is generally acceptable when compared against an intervention arm, and trial arms are balanced with respect to standards of care. Our study design will not withhold standard of care for any medical treatment or intervention, in alignment with WHO-VCAG guidance.
Criteria for Discontinuing or Modifying Allocated Interventions:
All participation is completely voluntary, and community members can withdraw at any time with impunity. Study staff may terminate subject participation at any time during the trial, as necessary, if a subject no longer meets the inclusion criteria and/or based on adverse event (AE) and/or severe adverse event (SAE) clinical assessment.
Strategies to Improve Adherence to Interventions:
Because trial interventions are implemented at the camp level, not the shelter level, we do not anticipate any issues of participant intervention adherence. The study team will ensure that interventions are deployed at the correct coverage per camp during each implementation phase. Throughout the entomological monitoring, the study team will also perform periodic spot checks that no modifications have been made to pit latrines and intervention re-treatment may be considered, as appropriate.
Relevant Concomitant Care Permitted or Prohibited During the Trial:
Our study design will not withhold standard of care for clinical management or intervention of any medical condition in either trial arm. Camps assigned to either trial arm will continue to receive balanced humanitarian aid and access to healthcare and other medical services.
Outcomes: See "Outcome Measures"
Participant Timeline:
From enrolment to the end of the cross-over trial intervention phases.
Sample Size:
Baseline diarrhoeal disease prevalence data collected during the trial census was used to calculate sample sizes, targeting a 20% relative reduction in diarrhoeal incidence (episodes per person-time), with 80% power and FWE = 0.05 across three primary comparisons, for two pre-specified age cohorts (≤5 years old and ≥15 years old). Diarrhoeal disease prevalence was 7.20% (898/12476) and 4.55% (1231/27039) in ≤5 years old and ≥15 years old, respectively, at trial baseline (MENTOR Initiative, unpublished data). The design comprises a 3-period cross-over: two implementation periods (phases I-II) and one combined-intervention period (phase III), each consisting of repeated fortnightly measurements. Recruitment targets allowed for 20% loss-to-follow-up. Under these assumptions, the minimum fortnightly sample per camp in implementation phase I-II is estimated to be 650 and 325 individuals for the ≤5 cohort and 1202 and 601 individuals for the ≥15 cohort, in the intervention and control camps, respectively; in implementation phase III we will enroll 627 for the ≤5 cohort and 1147 for the ≥15 cohort in all camps. In the absence of an intervention effect, in phase I-II, per phase, this is expected to yield 608 cases in each intervention camp and 304 cases in each control camp in the ≤5 cohort and 710 cases in each intervention camp and 355 cases in each control camp in the ≥15 cohort. In phase III, this is expected to yield 587 cases per camp in the ≤5 cohort and 678 cases per camp in the ≥15 cohort.
Latrine level density of filth flies will be estimated using quadrat sampling. Targeting a mean difference in density with an effect size of 0.80, an intraclass correlation coefficient (ICC) of 0.15, a power of 80%, and a Bonferroni corrected α = 0.0167, a total of 43 latrines will be sampled per trial arm. In implementation phase I-II, each intervention camp will target 43 latrines; each control camp will target 22. In implementation phase III, each camp will target 22 latrines.
Recruitment:
All participants will be recruited from the 4 refugee camps (Doro, Kaya, Yusuf Batil and Gendrassa) in Maban County, Upper Nile State, South Sudan. Prior to trial commencement, a tiered and iterative approach will be taken to sensitize the communities to project objectives. Letters of introduction will be sent to each camp manager and local community leaders to seek permission to conduct the study. Camp managers, community leaders and local health staff will be invited to sensitization sessions, where the study team will inform them of the project, explain the procedures and timelines and address any questions or concerns. Due to the changing dynamics of this humanitarian emergency, the research team will continually re-assess camp security and accessibility before trial commencement and throughout each implementation phase. Because new issues can arise at any point, project staff will regularly attend camp meetings throughout the baseline year and implementation years to address any problems and to provide updates of study progress.
During trial baseline, project staff will perform a shelter-to-shelter census, where every shelter and pit latrine will be mapped with a Global Positioning System (GPS) and camp members will be informed of the study, procedures and timelines. Recruitment of participants ≤5 years old or ≥15 years old for the post-intervention cohorts will also occur shelter-to-shelter during the baseline census survey. Cohort recruitment per camp will continue until we reach our projected sample size, therefore, we anticipate no difficulty achieving recruitment targets. In each subsequent fortnightly survey, cohort monitoring will occur shelter-to-shelter by our trial clinical team. Our sample size calculation for both cohorts assumes a 20% loss-to-follow-up across 6 months, with individuals considered "lost-to-follow-up" if they miss at least 3 consecutive visits per implementation period. Because displaced populations can be mobile and the emergency in South Sudan is dynamic, to address issues of participant attrition between project years, each cohort will be re-enrolled at the beginning of each implementation phase. Despite the instability of this setting, we do not anticipate significant participant loss-to-follow-up per monitoring period, since the MENTOR Initiative has conducted several prior clinical trials in these camps, establishing strong local partnerships and a precedent for high study participant retention. Additional eligible individuals will be recruited per cohort, if we encounter greater than anticipated loss-to-follow-up.
Methods: Assignment of Interventions Allocation and Implementation:
Because this is an interventional non-randomized cross-over trial design, sequence generation and concealment for participant randomization are not required. Intervention implementation will occur after completion of trial baseline analyses and verification of underlying assumptions of study power. During implementation phase I, Doro camp will receive TRAP and Kaya camp will receive LARV. During implementation phase II, Doro camp will receive LARV and Kaya camp will receive TRAP. During implementation phase III, Doro and Kaya camps will receive both LARV and TRAP. Participants will be recruited to 2 prospective cohorts for monitoring all-cause diarrhoeal case incidence. Camps have been pragmatically allocated to study arms, to minimize intervention contamination from filth fly dispersal range.
Methods: Assignment of Interventions Blinding and Emergency Unblinding:
Due to the obvious nature of the intervention implementation, study participants and project staff will not be blinded to camp allocation. All data analyses will be performed blinded; only 1 study PI will have access to the unblinded camp codes. Blinded data will be shared with the Data Safety and Monitoring Board (DSMB) who can request unblinding, if they have concerns. If an AE may be related to either intervention and non-emergency unblinding is considered, this will follow recommendations outlined in pre-specified standard operating procedures (SOPs). Emergency unblinding will be considered in instances of a suspected, unexpected SAE to either intervention, as judged by site project staff, following recommendations outlined in pre-specified SOPs.
Methods: Data Collection, Management and Analysis:
Baseline Census At trial baseline, a census questionnaire will be used to collect demographic details on camp residents, including number of people living in each shelter, number of children, participants' ages, family assets (including owning animals within the camp location, as a composite metric for socio-economic status; SES), shelter construction materials, condition and size, distance to nearest health facility, whether cooking is done inside or outside of the shelter, WASH access, distance to nearest pit latrine and latrine type, and distance to nearest domestic waste site. Census data will be used to randomly select eligible individuals for the 2 active monitoring cohorts in each camp and to estimate intervention quantities required. Every shelter, pit latrine and domestic waste disposal site per camp will be mapped with a GPS.
Epidemiological Monitoring To evaluate the effectiveness of LARV, TRAP, or LARV + TRAP in reducing the number of cases of all-cause diarrhoea, epidemiological data will be collected from 2 sources during baseline and each of 3 implementation phases. We will collect information on 1) diarrhoea incidence from local health facilities; and 2) shelter level diarrhoea incidence using fortnightly surveys. Passive surveillance data on diarrhoeal cases from camp residents will be obtained from the nearby health facilities. Permission will be sought from these facilities during trial baseline, and fortnightly data will be collected during each 6-month implementation phase.
For shelter-level diarrhoea monitoring, camp residents will be recruited into 2 cohorts to be followed through each implementation phase. Study participants will be assigned to 1 of 2 cohorts: ≤5 years old, and ≥15 years old. The rationale for monitoring both age categories is as follows. Most diarrhoeal studies focus on younger age groups as incidence and mortality caused by diarrhoeal illness is much higher in children ≤5 years. This sentinel age category may suffer from increased transmission from poor sanitation and hygiene practices, and young children with worse nutritional status and reduced immunity are more likely to present with illness of greater clinical severity and are therefore more likely to be reported. Monitoring of this high prevalence cohort will allow us to determine if different combinations of interventions can reduce incidence amongst this vulnerable group experiencing multiple modes of diarrhoeal pathogen transmission. We will also collect diarrhoeal incidence data from camp residents that are ≥15 years old. This older cohort is of particular importance as they may be assumed to have improved hygiene practices; therefore, mechanical transmission of diarrhoeal pathogens by flies may be proportionally greater in this age group than amongst younger children. Monitoring of diarrhoeal incidence in this population will also inform how these different vector control interventions can be used to interrupt disease transmission to military servicemen. In most shelters, we will strive to recruit multiple members of the family - 1 participant in the ≤5 years old cohort and at least 1 participant for the ≥15 years old cohort. Normally, this will be the youngest child and a parent or other adult(s) in the household.
Surveys will be completed fortnightly shelter-to-shelter, by surveyors trained in obtaining consent and performing accurate data collection in tablet devices using Kobo Toolbox. The study questionnaire will include items adapted from the UNICEF questionnaire for active diarrhoeal surveillance. The UNICEF questionnaire uses the WHO definition of diarrhoea (three or more loose or watery stools in a day) while allowing 2 weeks as the recall period: "Has (NAME) had diarrhoea in the last 2 weeks, that is three or more loose or watery stools in a day?". This provides coverage of reported diarrhoeal cases for the entire month. Visual tools will be incorporated within the questionnaire to establish if the participant is experiencing diarrhoeal illness and to improve accuracy in reporting. For the ≥15 years old cohort, the survey will contain the Bristol Stool Chart, whereas the ≤5 years old cohort will use the Amsterdam Scale. Each of these scales is specifically designed as visual aids for diarrhoea recognition for these age groups. For each shelter included in the study, basic socio-demographic information will also be collected, along with the occurrence of any AEs or SAEs or other related health symptoms, such as fever, headache, or any other illness that has affected the study participant or any other member of the household within the last 2-week period. Study participants will also be asked about intervention perceived effectiveness and other questions pertaining to WASH access and behaviours. A cohort participant (child ≤5 years or person ≥15 years) will be considered "lost-to-follow-up" if they miss at least 3 consecutive visits per implementation period.
Entomology Monitoring in Sentinel Sites Population Density To evaluate the effectiveness of LARV, TRAP or LARV + TRAP in reducing filth fly density at the shelter and latrine level, we will collect data about fly densities every two weeks throughout each implementation phase. In each of the four camps, sentinel sites will be randomly selected and maintained throughout each 6-month study cycle, for all entomological monitoring activities.
Data will be collected at the latrine level using a Scudder grill. Every two weeks, in each of the sentinel sites per camp, project staff will monitor fly ground resting density around pit latrines. At each latrine, staff will place a Scudder grill in the middle of the latrine; disturbed flies will be allowed 30 seconds to resettle onto the grill, then a photo of the flies will be taken to later morphologically identify fly family, count and establish an approximate density value. The counts will be repeated in similar weather conditions and around the same time of the day.
In each sentinel site per camp, data will also be collected at the shelter level using iCatchi® ProMax fly traps. Twice monthly, during each 6-month implementation phase, one iCatchi® ProMax fly trap will be placed in each of the sentinel sites. It will be baited and put in a fixed outdoor location (<5 m) amongst consenting shelters for a period of 30 mins. Each trap will be collected, taken to the field entomology laboratory, and the contents will be removed carefully for sample extraction, processing, morphological speciation recording, counting, and immediate preservation in a buffer solution. Samples will be placed directly into field storage at -20oC for future laboratory analysis and sub-studies.
Shelters and pit latrines in each of the entomological sentinel sites will be eligible for inclusion in the entomological monitoring, provided they received the interventions that they were assigned at the beginning of that implementation phase.
Active Ingredient Efficacy Monitoring:
To monitor Dimilin® (LARV) residual efficacy, monthly pyramidal emergence traps will be used to sample emergent flies in a random subset of 40 treated pit latrines across sentinel sites, per trial arm assigned to this intervention.
To assess any changes in filth fly insecticide susceptibility following intervention deployment, resistance monitoring will be performed. At baseline and at 3- and 6-months post-implementation, in each 6-month implementation phase, wild filth fly populations will be sampled from all 4 camps and exposed to diflubenzuron in larval bioassays. Adult emergence inhibition concentrations required to prevent successful adult emergence in 50%, 95% and 99% of exposed larvae (EI50, EI95 and EI99) will be established at trial baseline and used to monitor changes in diflubenzuron resistance post-intervention.
For larval bioassays, gravid filth fly populations will be allowed to oviposit onto organic rearing media (created by mixing 1 L of water and 2 L of dry rearing media). Per bioassay, groups of twenty, 3- to 5-day old, 2nd instar larvae will be exposed to serial dilutions of diflubenzuron spiked into rearing media (3 replicates will be performed per concentration with an untreated control run in parallel). Larvae will be left on media at 23 ± 2°C, 50% RH and 12L:12D photoperiod until all individuals have either emerged as adults or failed to complete development. The primary endpoint will be inhibition of adult emergence, defined as the proportion of exposed larvae that do not successfully emerge as viable adults. Larval mortality, pupation, non-emerged pupae and adult emergence will be recorded separately to distinguish mortality before pupation from failure of pupal / adult emergence.
Laboratory Analysis:
A series of sub-analyses will be conducted on fly samples collected during each 6-month intervention phase, as follows:
Field Laboratory Testing Viability of pathogens inside or on flies will be demonstrated through testing freshly trapped field fly samples for Escherichia coli (E. coli). Tests will be conducted on pooled (5) fly samples (pooled by morphologically identified species from each of the ten sampling sites). Each pooled fly sample will be crushed and diluted with PBS in separate cryotubes to create dilutions of 10^0 and a 10^(-2) for each fly pool. Each fly soup dilution will be pipetted onto separate pre-prepared EC plates, and cefotaxime will be plated onto a third plate, then incubated at 35oC for 24 hours, together with labeled negative control plates. E. coli presence, or not, will be carefully recorded and data input into an Excel spreadsheet. Any fly samples that are TNTC for E. coli at the 10^(-2) dilution will be re-tested.
UNLV Laboratory At least 200 fly samples collected during each of the three intervention phases, across the study arms will be processed for laboratory testing at UNLV. These will be immediately preserved in buffer solution in 2 ml Eppendorf tubes and then frozen and stored at -20oC in the field laboratory. A replicate of 200 flies sampled for each intervention phase will also be preserved and retained in freezer storage as a contingency sample.
At the end of each intervention phase, 200 of the preserved and frozen fly samples collected during that 6-month phase will be transported to UNLV by cold chain, for fly and pathogen identification during each wash-out period.
The laboratory analysis will include morphological identification of individual flies to species level, performed according to standard dichotomous taxonomic keys. Per trial 6-month period, 200 randomly sampled flies (split equitably across camps / trial arms) will be selected for molecular analysis. A total of 600 flies will be analysed over the course of the study. Of these, a subset of at least 100 flies will be analysed using a custom TaqMan Array Card (TAC), initially in order to identify the most prevalent pathogens in the samples. Following the completion of TAC analysis on the subset of 100 flies, a further 500 flies will be analysed by quantitative polymerase chain reaction (qPCR) for the most prevalent targets identified by TAC.
Data Management:
Epidemiological and entomological data will be entered into electronic forms on smartphones or tablets installed with Kobo Toolbox. The data will be stored on a secure server located at the MENTOR Initiative in-country office. All data management and analyses will be performed using StataNow/MP 19.5 and RStudio v. 2025.09.0-387 or higher.
Data Quality and Control (QC). SOPs for data collection will be developed, and study data collectors will be appropriately trained to ensure rigorous data collection. QC will be conducted by a supervisor who will monitor the performance of the data collectors.
Statistical Methods - Outcomes:
Because intervention allocation is not fully randomized, direct post-intervention comparisons between trial arms may be confounded by time-invariant between-camp differences and by secular or seasonal temporal changes. To account for these potential sources of confounding, the primary analysis will use a difference-in-differences (DID) framework that contrasts within-camp changes from baseline against contemporaneous changes over the same periods in the control camps. All analyses will follow the intention-to-treat (ITT) principle and will be conducted separately in two pre-specified age cohorts (≤5 years old and ≥15 years old). The primary epidemiological outcome is diarrhoeal incidence (episodes per person-time), modelled using generalized linear mixed models (GLMMs) with a log link and log person-time offset. Two covariate sets were defined a priori. The core model (primary analysis) will comprise: (i) camp fixed effects, (ii) phase fixed effects, (iii) armxphase interaction (the DID term), (iv) baseline diarrhoeal prevalence (measured pre-implementation within each age cohort), and (v) survey round to account for seasonality. Exploratory covariates (e.g. socio-economic status, population density and pit latrine density) will be added only in pre-specified sensitivity models to evaluate robustness and potential effect modification, not to redefine the primary estimand. Random intercepts will account for shelter clustering and repeated measurements within participants.
The primary estimands will be arm-specific DID incidence rate ratios (IRRs) defined as the ratio of the within-arm change in incidence from baseline to each post-baseline phase relative to the contemporaneous change from baseline in the control arm (ratio of IRRs). Primary contrasts will be specified as: (i) LARV vs. control and (ii) TRAP vs. control, each estimated as baseline-to- phase I-II DID effect (time-averaged across phases I-II via linear contrasts of armxphase interaction terms), and (iii) LARV + TRAP vs. control, estimated as the baseline- to-phase III DID effect (contemporaneous in phase III). Because the unit of intervention assignment is camp and with only four camps included, statistical inference for the primary contrasts will rely on small cluster approaches (e.g. wild cluster bootstrap at the camp level and / or sensitivity analyses using camp-phase aggregated models), rather than asymptotic cluster-robust standard errors. We will report two-sided 95% CIs and p-values for each primary contrast; family-wise error (FWE) rate across the three primary comparisons will be controlled using a Holm step-down procedure.
Following baseline data collection, balance between trial arms will be summarized descriptively and quantified using standardized mean differences (SMD) for each baseline covariate, emphasizing magnitude rather than significance testing. The primary DID GLMM includes camp fixed effects and baseline diarrhoeal prevalence, thereby controlling for time-invariant between camp differences and baseline outcome level. As pre-specified analyses, we will (i) adjust for baseline covariates showing meaningful imbalance (e.g. SMD > 0.10), and (ii) if indicated by pre-intervention outcome patterns, include camp-specific linear time trends to relax the parallel trends assumption.
The secondary epidemiological outcome will compare fortnightly diarrhoea prevalence (i.e. the proportion of participants reporting ≥1 diarrhoea episode during the preceding 14-day recall period) between arms using GLMMs with the same camp and period structure as the primary DID framework. We will fit log binomial GLMMs with comparable fixed / random effects and covariates to estimate prevalence ratios (PRs); if convergence fails, we will use Poisson GLMMs with robust variance to obtain PRs. Period-standardized marginal prevalences and PRs will be obtained from model-based predictions using marginal standardization over the observed covariate distribution. Key secondary contrasts include LARV vs. TRAP (time-averaged over phases I-II via linear contrasts of the phase-specific log(PR) terms) and an assessment of departure from multiplicativity for LARV + TRAP in phase III relative to the separate LARV and TRAP effects, implemented as a pre-specified interaction test on the multiplicative scale using effects defined relative to the control condition.
For entomological outcomes, latrine and shelter level fly densities will be analysed using DID GLMMs with log link and log exposure offsets (trapping effort), including camp (cluster) effects, phase effects and armxphase interactions. Effects will be reported as DID rate ratios (ratio of IRRs) and phase standardized using model-based marginal predictions. Filth fly enteric pathogen species composition will be modelled with multinomial logistic mixed models, including phase effects, and armxphase interactions; when over-dispersion is evident, Dirichlet multinomial mixed models will be used. We will report global armxphase tests of compositional differences and secondarily, species-specific contrasts with false discovery rate (FDR) control within phase. Enteric pathogen burden will be modelled with GLMMs, with an offset for the number of flies processed, to estimate pathogen burden per fly. AMR prevalence (any marker detected) will be analysed with log-binomial GLMMs (or Poisson GLMMs with robust standard errors as a fallback), reporting phase-standardised DID PRs.
For supporting laboratory bioassays, including insecticide susceptibility and bioefficacy monitoring, GLMMs will include appropriate random effects for test replicate and dose, and fixed effects for trial arm, phase and armxphase interaction to estimate temporal trends relative to baseline.
Inclusion Criteria:
Exclusion Criteria:
richard.allan@mentor-initiative.org+44 796193670
louisa.messenger@unlv.edu+1 702-449-3295
Bunj, Maban County, South Sudan
mohamed.sheikh@mentor-initiative.org+211 925219599
sajjad.ahmad@mentor-initiative.org