Thickstun Charles, Lukole Eliud, Mosha Jacklin F, Manjurano Alphaxard, Kleinschmidt Immo, Mosha Franklin W, Charlwood Jacques Derek, Rowland Mark, Jolly Ann, Zwerling Alice, Protopopoff Natacha, Kulkarni Manisha
School of Epidemiology and Public Health, University of Ottawa, Ottawa, ON, Canada.
Department of Parasitology, National Institute for Medical Research, Mwanza Medical Research Centre, Mwanza, Tanzania.
Malar J. 2025 Jun 9;24(1):184. doi: 10.1186/s12936-025-05438-y.
The presence of a community effect in cluster randomized trials of malaria vector control interventions has led to the implementation of "buffer zones" around clusters to limit the potential for contamination between interventions. No consensus has been reached on how large these buffers need to be to encapsulate the effect.
Nested within a phase-III cluster randomized malaria vector control trial in Northwest Tanzania, this study aims to determine the presence and spatial range of community effects from long-lasting insecticidal net (LLIN) and indoor residual spraying (IRS) interventions on household-level malaria infection in trial clusters four months post-intervention. Effective spatial range estimates of intervention community effects were compared to the 300m buffer distance implemented to limit intervention spillover between clusters in the trial. Geographically-weighted adjusted odds of malaria infection in children aged 0.5-14 years were determined four months post community-level intervention with a randomized allocation comprising one of two LLIN products (Olyset LN: 1000mg/m permethrin or Olyset Plus LN: 400 + permethrin 800mg/m) with either IRS (Actellic300CS: 1000mg/m micro-encapsulated pirimiphos-methyl) or no IRS. Robust semivariances were calculated for each of 48 intervention clusters and fit to semivariogram models by Weighted Least Squares.
6440 children from 2785 households were included in the geographically-weighted logistic regression. Prevalence of Plasmodium falciparum infection was 45.9% in the study population. Twenty (20) clusters had significant residual effect ranges, 13 of which were fit to Sine Hole Effect models, indicating periodicity in the study area. Effective range estimates for the study area had a median value of 1210 m (IQR: 958-1691). Clusters with IRS had a higher median range value: 1535 m (IQR: 976-3398) than those without IRS: 1168m (IQR: 829-1504).
Significant semivariogram model range estimates extended beyond the trial buffer sizes by a median average of 868 m in LLIN intervention clusters and 1235 m for IRS clusters. This presents a contamination, or spillover, potential for all trialed intervention types that may reduce the statistical power to detect difference between trial arms. Future studies should consider the ranges of intervention effects and contamination potential between trial arms when designing buffer areas.
疟疾媒介控制干预措施的整群随机试验中存在社区效应，这导致在各群组周围设置“缓冲区”，以限制不同干预措施之间的潜在污染。对于这些缓冲区需要多大才能涵盖该效应，尚未达成共识。
本研究嵌套于坦桑尼亚西北部的一项三期整群随机疟疾媒介控制试验中，旨在确定在干预四个月后，长效驱虫蚊帐（LLIN）和室内滞留喷洒（IRS）干预措施对试验群组中家庭层面疟疾感染产生的社区效应的存在情况及空间范围。将干预社区效应的有效空间范围估计值与为限制试验中各群组间干预溢出而设置的300米缓冲距离进行比较。在社区层面干预四个月后，确定0.5 - 14岁儿童疟疾感染的地理加权调整比值比，随机分配包括两种LLIN产品之一（Olyset LN：1000mg/m氯菊酯或Olyset Plus LN：400 + 800mg/m氯菊酯），同时搭配IRS（Actellic300CS：1000mg/m微囊化甲基嘧啶磷）或不使用IRS。为48个干预群组分别计算稳健半方差，并通过加权最小二乘法拟合到半变异函数模型。
2785户家庭的6440名儿童纳入地理加权逻辑回归分析。研究人群中恶性疟原虫感染率为45.9%。20个群组具有显著的残余效应范围，其中13个符合正弦孔效应模型，表明研究区域存在周期性。研究区域的有效范围估计值中位数为1210米（四分位间距：958 - 1691）。采用IRS的群组中位数范围值更高：1535米（四分位间距：976 - 3398），高于未采用IRS的群组：1168米（四分位间距：829 - 1504）。
显著的半变异函数模型范围估计值在LLIN干预群组中比试验缓冲区大小平均中位数超出868米，在IRS群组中超出1235米。这表明所有试验干预类型都存在污染或溢出的可能性，可能会降低检测试验组间差异的统计效力。未来研究在设计缓冲区时应考虑干预效应范围和试验组间的污染可能性。
