Williamson Tyler, Ravani Pietro
O'Brien Institute of Public Health, Department of Community Health Sciences, Cumming School of Medicine, University of Calgary, Calgary, Alberta, Canada.
Alberta Children's Hospital Research Institute, Cumming School of Medicine, University of Calgary, Calgary, Alberta, Canada.
Nephrol Dial Transplant. 2017 Apr 1;32(suppl_2):ii84-ii90. doi: 10.1093/ndt/gfw341.
Marginal structural models are a multi-step estimation procedure designed to control for the effect of confounding variables that change over time, and are affected by previous treatment. When a time-varying confounder is affected by prior treatment standard methods for confounding control are inappropriate, because over time the covariate plays both the role of confounder and mediator of the effect of treatment on outcome. Marginal structural models first calculate a weight to assign to each observation. These weights reflect the extent to which observations with certain characteristics (covariate values) are under-represented or over-represented in the sample with the respect to a target population in which these characteristics are balanced across treatment groups. Then, marginal structural models estimate the outcome of interest taking into account these weights. Marginal structural models are a powerful method for confounding control in longitudinal study designs that collect time-varying information on exposure, outcome and other covariates.
边际结构模型是一种多步骤估计程序，旨在控制随时间变化且受先前治疗影响的混杂变量的效应。当一个随时间变化的混杂因素受到先前治疗的影响时，用于控制混杂的标准方法就不合适了，因为随着时间的推移，协变量既起到混杂因素的作用，又起到治疗对结局效应的中介作用。边际结构模型首先计算分配给每个观察值的权重。这些权重反映了具有某些特征（协变量值）的观察值在样本中相对于目标人群的代表性不足或过度代表性的程度，在目标人群中这些特征在治疗组之间是平衡的。然后，边际结构模型在考虑这些权重的情况下估计感兴趣的结局。边际结构模型是纵向研究设计中控制混杂的有力方法，该设计收集有关暴露、结局和其他协变量的随时间变化的信息。
