A recent cohort study revealed a positive correlate between major structural birth defects in infants and a certain medication taken by pregnant women. To draw valid causal inference, an outstanding problem to overcome was the missing birth defect outcomes among pregnancy losses resulting from spontaneous abortion. This led to missing not at random since, according to the theory of "terathanasia", a defected fetus is more likely to be spontaneously aborted. Other complications in the data included left truncation, right censoring, observational nature, and rare events. In addition, the previous analysis stratified on live birth against spontaneous abortion, which was itself a post-exposure variable and hence did not lead to a causal interpretation of the stratified results. In this paper we aim to estimate and provide inference for the causal parameters of scientific interest, including the principal effects, making use of the missing data mechanism informed by "terathanasia". The rare events with missing outcomes led to multiple sensitivity analyses where the causal parameters can be estimated with better confidence in each setting. Our findings should shed light on how studies on causal effects of medication or other exposures during pregnancy may be analyzed using state-of-the-art methodologies.
翻译:最近的一项组群研究揭示了婴儿在出生方面的重大结构性缺陷与孕妇服用的某种药物之间的正相关关系。为了得出有效的因果推断,一个有待克服的问题是自发堕胎造成的怀孕损失中缺少出生缺陷的结果。这导致并非随机失踪,因为根据“畸形”理论,一个叛变的胎儿更有可能自发堕胎。数据中的其他并发症包括左侧脱节、右审查、观察性质和罕见事件。此外,先前对活产与自发堕胎(这本身就是一种接触后变数,因此没有导致对分层结果的因果解释。在这份文件中,我们的目标是利用“畸形”所通报的缺失数据机制,估计和推断科学利益因果关系参数,包括主要影响。缺损数据机制的缺损结果导致多重敏感性分析,从而可以更有信心地估计因果参数。我们的调查结果应说明如何利用州法分析孕期药物或其他照射的因果影响研究。