Potential confounding by exposure history and prior outcomes - An example from perinatal epidemiology
Potential confounding by exposure history and prior outcomes - An example from perinatal epidemiology
复制标题
DOI：
10.1097/ede.0b013e31812001e6
复制
发表时间：
2007-09-01
期刊：
影响因子：
5.4
通讯作者：
Heagerty, Patrick J.
中科院分区：
文献类型：
--
作者：
Howards, Penelope P.;Schisterman, Enrique F.;Heagerty, Patrick J.
先前的妊娠结果，如自然流产和早产，通常可以预测未来的妊娠结果。因此，许多研究人员对生殖史进行了调整。虽然这种调整可能适用于预测模型，但当目标是获得暴露对疾病影响的无偏估计时，它不一定适用。生殖史似乎符合传统的混杂标准，因为它不太可能在暴露和当前结局之间的因果关系上，通常与当前结局相关，也可能与暴露相关。然而，生殖史是否是混杂因素取决于其与暴露和当前结果相关的根本原因。因此，评估混杂的传统方法往往是不够的。有向无环图（DAG）可用于评估复杂的情况下，当研究问题是明确定义的暴露，结果和效果估计的利益。当生殖史影响未来的暴露时，需要特别注意。我们使用5个DAG来说明生殖史和当前结果之间的可能关系。我们评估每个DAG的混杂因素，并确定适当的分析技术。我们提供了一个数字的例子，使用围产期合作项目的数据。关于生殖史是否应该包括在模型中没有单一的答案;决定取决于研究问题和潜在的DAG。
Prior pregnancy outcomes, such as spontaneous abortion and preterm birth, are often predictive of future pregnancy outcomes. Therefore, many researchers adjust for reproductive history. Although this adjustment may be appropriate for a predictive model, it is not necessarily appropriate when the goal is to obtain an unbiased estimate of the effect of exposure on disease. Reproductive history may seem to meet the conventional criteria for confounding because it is unlikely to be on the causal pathway between exposure and current outcome, is often associated with current outcome, and may be associated with exposure as well. However, whether reproductive history is a confounder or not depends on the underlying reason for its associations with exposure and current outcome. Thus, conventional methods for assessing confounding are often inadequate. Directed acyclic graphs (DAGs) can be used to evaluate complex scenarios for confounding when the research question is clearly defined with respect to the exposure, the outcome, and the effect estimate of interest. Special care is required when reproductive history affects future exposure. We use 5 DAGs to illustrate possible relations between reproductive history and current outcome. We assess each DAG for confounding, and identify the appropriate analytic technique. We provide a numeric example using data from the Collaborative Perinatal Project. There is no single answer as to whether reproductive history should be included in the model; the decision depends on the research question and the underlying DAG.