Study site
The study was implemented in ten communities (Njayel, Banni Kunda, Temanto, Bolibana, Fula Mori Bochi, Madina Samba Sowe, Njum Bakary, Sare Demba Dardo, Sare Biram and Tabajang) in the Upper River Region (URR) of the Gambia (Fig. 1). The region has a population of over 250,000 inhabitants and has a distinctly seasonal pattern of malaria transmission, where the most cases occur between June and December, peaking directly following the annual rains between June and September29. The main vector species is Anopheles gambiae sensu latu30,31. During the 2021 season, the mean nightly Anopheles gambiae s.l. and Anopheles funestus catch rate per house in the study communities using Centres for Disease Control light traps rate was 0.97 (standard deviation 1.41) (Supplementary File STable 1). The peak temperature during the transmission season was 35.2 °C in 2021 (Supplementary File STable 2). A region-wide distribution drive of long lasting insecticide treated nets occurred in 2019. For the first time in July 2021, The NMCP extended the upper eligible age for SMC from 4 to 9 years. Eligible children aged 0–9 years could receive up to four rounds of monthly SMC with sulphadoxine-pyrimethamine (SP) and amodiaquine (AQ) administered between August and November 2021 by community outreach teams.
Study design
This prospective observational cohort study was nested within a larger trial measuring the effectiveness of control interventions on malaria infection and transmission, the INDIE Trial (P. falciparum Infection Dynamics and Transmission to Inform Elimination, clinicaltrials.gov reference NCT04053907) and used data collected between 26th July 2021 and 12th January 2022. Three communities (Njum Bakary, Sare Demba Dardo and Sare Biram) were randomly assigned to receive 3 monthly rounds of mass drug administration with dihydroartemisinin-piperaquine between April and June 2021. In two communities (Madina Samba Sowe and Tabajang) inhabitants were screened weekly for fever—‘fever screen and treat (FSAT)’, those with temperatures above 37.5 °C or a history of fever in the last week received a rapid diagnostic test (RDT) for malaria and positive cases were treated with artemether-lumefantrine. Three communities (Banni Kunda, Temanto and Njayel) were assigned to monthly screening of all residents with ultrasensitive-RDTs, and positive cases treated with artemether-lumefantrine—‘mass screen and treat’ (MSAT). The remaining two communities (Bolibana and Fula Mori Botchi) were assigned to the control group and received standard control interventions. A programme of community case management of malaria was concurrently initiated in all communities. From 26 July 2021, a clinician and support worker (nurse and community health worker) based in each community managed all suspected malaria cases (passive case detection—PCD).
Study procedures
A series of community-based sensitisation activities took place during the INDIE baseline year (2019) to inform community members about the study and its aims, and all inhabitants were invited to take part. Residents who provided informed consent were enroled during a baseline dry season survey in April-May 2021. Cross-sectional surveys covering all enroled community members were carried out on a rolling basis every 8 weeks during the transmission season. Blood samples collected from the enroled population during all surveys were analysed at the MRC Unit The Gambia (MRCG) in Basse by quantitative polymerase chain reaction (qPCR varATS) to detect malaria infections. Study questionnaires were administered during the surveys to all participants and collected address details and GPS coordinates, demographics including age, gender and ethnicity, symptoms of illness, care seeking and any treatments received for each participant and use of insecticide-treated bed nets the night prior to the survey. Identical data were collected from malaria patients identified by PCD visits although diagnosis was done by RDT only. The dates and number of all SMC rounds received for each child under 10 years were collected from SMC cards or by caregiver report. SMC data were collected in a survey after the last (November) round of SMC, and at two additional capture points: during a final cross-sectional survey in January 2022 and following a review of the SMC database by SS and AN, in a dedicated SMC mop-up survey to address data queries and collect data on additional study children not met in previous surveys (10th–15th March 2022). Entomological surveillance was conducted during the 2021 study period in 6 randomly selected households per community. In each household, Centre for Disease Control (CDC) light traps were hung in a sleeping room for three consecutive nights per month; caught mosquitoes were examined to determine species, sex and gonotrophic status (fed status and gravidity).
Study outcomes
All data from PCD visits, cross-sectional and entomological surveys were collected onto secure handheld devices and stored, cleaned and accessed via a password protected REDcap® server (Vanderbilt University, Nashville Tennessee) hosted at the MRC LSHTM Research Unit in Fajara. Incidence of clinical malaria was defined as the number of passively detected cases per 100 person-months. For each clinical case, 2 weeks of follow up were removed from the denominator. Malaria prevalence was defined as the percentage of participants sampled who were qtPCR-positive during a late season survey (27th September to the 27th November 2021). qtPCR-positive participants sampled during this survey were categorised into high- and low-density infections by age group. High-density infections were those with density above the median of the natural log asexual parasite concentrations per µL blood for the age groups (0–4, 5–9, 10–15 and 16+ years).
Statistical analysis
The impact of SMC status on the incidence of clinical malaria in eligible children was assessed by fitting shared frailty cox models with gamma-distributed random effects specified for person to account for the possibility of repeated clinical episodes within the same person (failures). Where frailty models did not stabilise, robust errors were used instead. To account for ongoing malaria control interventions in each community and background community level risk of malaria, the models pre-specified fixed effects for village ID. Models also included fixed effects for the percentage of all visits to the participant at which an insecticide-treated bed net was used the night before and household ID.
To assess the association between household level coverage of SMC in eligible children and malaria burden in adolescents/adults 10+ years of age, we first defined household SMC coverage in two ways: (i) the percentage of children 0–9 years in each household who received at least 1 round of SMC and (ii) the mean number of rounds of SMC per child. Both definitions were implemented as continuous variables and classed into ‘low’, ‘moderate’ and ‘high’ groups thus: <25%: 25–79%, 80%+ (definition 1) and <0.3, 0.3–2 and >2 rounds/child (definition 2), respectively. The high group cut-offs (80% or >2 rounds) were fixed to be consistent with the optimum minimum level of coverage of child health programmes32,33,34. The low/medium cut-off values were varied for definition 1 in the range 0–55% in 5% increments and for definition 2 in the range 0–0.9 in 0.1 unit increments and implemented in household level models described in Supplementary File 1 STable 3. The models with low groups of <25% and 0.3 had the lowest Akaike and Bayesian Information Criteria values (AIC and BIC) and were thus selected as the final low-group cut-offs in all household level models.
The impact of household-level coverage of SMC as defined above on clinical malaria episodes in participants 10+ years of age was then assessed by fitting shared frailty Cox models to account for repeated clinical episodes as previously. The impact of household-level coverage of SMC on malaria prevalence in participants 10+ years of age was estimated using logistic regression models with a household-level random effect. We visualised parasite concentration distributions in participants aged 10+ years stratified by household SMC coverage by the generation of kernel density estimate plots and estimated the impacts of household SMC coverage on the prevalence of high-density infections in this group using logistic regression models with a household-level random effect. Equivalent data in children 0–4 and 5–9 years were limited by fewer malaria infections and smaller cell sizes, therefore the additional household-level impacts of SMC on eligible children themselves were assessed in all eligible children aged 0–9 years combined (STables 7, 8 Supplementary File). Where multilevel models failed to converge, we calculated household-level estimates of malaria burden in this age group and regressed these against household-level SMC coverage using a two-stage technique to adjust for individual-level covariates35,36. All individual and household-level models of impact of SMC were implemented in Stata version 17.0 (StataCorp, Texas USA).
All household-level models pre-specified fixed effects for village ID, the percentage of all visits to households at which inhabitants used an insecticide-treated bed net the night before, the household level prevalence of malaria at a baseline dry season survey in April-May 2021, age in years, the household ratio of children aged 0–9 years to participants aged 10+ years, the total number of household inhabitants and for prevalence models, the week of survey visit.
To assess the effect of spatial clustering of household-level impacts of SMC, an inverse weighted distance matrix was created describing the Euclidean distances (km) between all households in the study site. We calculated the mean incidence rates of clinical malaria per participant aged 10 years and older per enumerated household, and the mean incidence rate in children 0–9 years per household and attached additional features of SMC coverage, insecticide-treated bed net usage, compound size and age composition, and geolocation (latitude and longitude) to each household. Identical datasets were constructed with the overall prevalence of asymptomatic infection per household during the late-season survey for the same age groups. Visual examples of the resulting datasets are shown in Supplementary Fig. 2 (Supplementary File). The distance matrix was used to calculate global Moran indices and z-scores to test for spatial autoregression under the hypothesis that prior regression estimates of the association between household-averaged outcomes for incidence and prevalence of malaria and household SMC coverage were not influenced by estimates in nearby households. Models were adjusted for household level covariates as previously stated – percentage of all visits to households at which inhabitants used an insecticide-treated bed net the night before, the household level prevalence of malaria at a baseline dry season survey in April-May 2021, the household ratio of children aged 0–9 years to participants aged 10+ years and the total number of household inhabitants. Visualisation of spatial data, construction of distance matrices and spatial analysis was conducted using Seaborn, Geoplot, Geopandas and the Esda_Moran libraries in Python (Python Software Foundation, Python Language Reference, version 3.9).
Ethical approval
This study did not collect any new data or implement additional activities not previously covered by approvals already in place for the INDIE study. Ethical approval for the INDIE study was provided by the Government of The Gambia/The MRC Gambia Joint Ethics Committee, The Gambia, and the Ethics Committee of the LSHTM, UK (reference: 16642).
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Credit: Source link