Dron L 2021

論文

Dron L, Taljaard M, Cheung YB et al. (2021) The role and challenges of cluster randomised trials for global health. The Lancet Global Health, 9(5), e701-e710.

https://doi.org/10.1016/S2214-109X(20)30541-6

https://www.thelancet.com/journals/langlo/article/PIIS2214-109X(20)30541-6/fulltext

著者を確認

https://orcid.org/0000-0002-6657-2381

Introduction

Certain interventions are delivered at a population level or at a group level, and these interventions can result in changes to group behaviours, leading to large scale population-level effects.\(^3\)

The cluster RCT design (herein, cluster trial) is a specific trial design that is used to evaluate interventions delivered at a group level.\(^4\) In cluster trials, whole groups of structured collections of individuals or health system service delivery platforms, such as facilities, are randomly assigned to receive interventions, and these groups are referred to as clusters. Examples of clusters include communities, health clinics, or schools.

Note

クラスタ RCT はどちらか?

• 病院Aを介入群、病院Bを対照群とする
• 病院A内で半数を介入群、残りを対照群とし、病院B内で半数を介入群、残りを対照群とする

In contrast to the individual RCT, in which the group allocation of interventions is determined by randomisation of individual participants, cluster trials randomly assign interventions to a whole cluster of individuals. Interventions themselves can be administered at a cluster level (eg, mosquito egg traps\(^5\)) or at an individual level (eg, vaccinations\(^6\)).

There is an increasing popularity of the use of cluster trials in low-income and middle-income countries (LMICs).\(^8\)

LMIC: low-income and middle-income country

Because outcomes of individuals within the same cluster are correlated, standard methods for design and analysis of individual RCTs do not suffice for cluster trials.

Why cluster trials are often used in research set in LMICs

Cluster trials can be useful in LMIC settings for several reasons. Cluster randomisation can potentially reduce treatment contamination between intervention and control groups.\(^{4, 10, 27}\)

If participants were individually assigned to the control group, they could potentially receive food supplements intended for the intervention group, resulting in a partial effect of the intervention in the control group, with or without a lesser effect in the intervention group.

Note

contamination: 汚染。ランダム化比較試験では、対照群が介入の影響を受けることを contamination という。

By contrast, if the intervention can be delivered at the level of the individual, it might be preferable to conduct an individual RCT with a larger sample size, rather than by cluster randomisation.\(^{31}\) The delivery of an intervention at an individual level is particularly appropriate in circumstances in which there is only a modest risk of contamination, there is a large intracluster correlation (ICC), or the cluster sizes are large.\(^{27}\)

Note

intracluster correlation (ICC): クラスタ間の相関

External validity

Another reason for the popularity of cluster trials over individual RCTs is the perceived improved external validity (or generalisability).\(^{4, 32}\) This perception of improved external validity is particularly prevalent when interventions engage large cluster groups, such as health districts.

Note

external validity と同義語はどれか？

• generalisability
• popularity

Many individual RCTs are carried out with specific types of patients under controlled conditions, with strict inclusion and exclusion criteria (mechanistic individual RCTs) for improved internal validity, but often this internal validity is achieved at the expense of external validity (explanatory trials).

発展

Leyrat C, Morgan KE, Leurent B & Kahan BC (2018) Cluster randomized trials with a small number of clusters: which analyses should be used?. International journal of epidemiology, 47(1), 321-331. https://doi.org/10.1093/ije/dyx169