Maternal Smoking During Pregnancy and Childhood Obesity: linear association or threshold effect? An IPD Meta-Analysis

在最近的一篇最先进的论文中，Gillman MW和Ludwig DS(NEJM，2013)提出，7到10岁时肥胖概率的7%可以用母亲怀孕期间吸烟来解释。然而，这种关于宫内尼古丁暴露与后代超重之间因果关系的经验证据受到了质疑，因为潜在的残留混杂。对可能的残留混淆的主要关切是基于这样的观察，即在宫内或怀孕后暴露于父亲或二手烟的儿童也有超重的风险增加，尽管比母亲吸烟的风险小。这是线性剂量效应关系的反映吗？或者，这两种影响都反映了未知家庭因素的残余混淆？在母亲或父亲怀孕期间吸烟的新生儿之间，新生儿头发中的可替宁浓度有一定的重叠，因此可能出现线性剂量效应关系。然而，已报道的关于孕妇吸烟的剂量效应的数据是模棱两可的：一些数据表明，与小剂量相关的阈值效应急剧增加，而另一些数据则表明两者之间存在线性联系。这些模棱两可的发现可能反映了由于按剂量类别报告的影响而产生的伪影。然而，大多数研究的原始数据都是通过吸烟数量来确定母亲吸烟的情况。因此，对这些数据的IPD荟萃分析将有助于阐明怀孕期间母亲吸烟的剂量效应的真实性质。16项研究的作者表达了他们对一项合作的IPD研究的兴趣，该研究允许在超过30万对母婴中研究这种关联。这些数据将为理解怀孕期间母亲吸烟与子女肥胖之间的潜在因果关系提供线索。在线性关联的情况下，观察到的孕期父亲吸烟的额外影响可能可以用剂量效应来解释。在与少量香烟有关的阈值效应急剧增加的情况下，母亲和父亲吸烟的影响更有可能反映与共同家庭环境有关的残余混淆。

In a recent state of the art paper Gillman MW and Ludwig DS (NEJM, 2013) suggested that 7% of the probability of obesity at 7 to 10 years of age could be explained by maternal smoking in pregnancy. This empirical evidence for a causal association between intrauterine exposure to nicotine and overweight in the offspring has been questioned, however, because of potential residual confounding. Major concerns regarding possible residual confounding were based on the observation that children who were exposed to paternal or second-hand smoking while in utero or after pregnancy also had an increased risk of being overweight, although smaller than for maternal smoking. Is this a reflection of a linear dose effect relation? Or are both effects are reflection of residual confounding by unknown family factors? A linear dose effect relation might appear possible in face of some overlap in cotinine concentrations in the newborns hair between newborns exposed to maternal or paternal smoking in pregnancy. The reported data on the dose effect of maternal smoking in pregnancy are equivocal, however: some suggesting a threshold effect with a steep increase related to small doses while others suggest a linear association. These equivocal findings may reflect artefacts due to reporting effects by dose categories. The original data of most studies, however, ascertained maternal smoking exposure by number of cigarettes smoked. An IPD meta-analysis of these data will therefore allow to elucidate the true nature of the dose effect of maternal smoking in pregnancy. The authors of 16 studies have expressed their interest in a collaborative IPD study allowing to study the association in more than 300 000 mother infant pairs. These data will provide a clue for understanding potential causality of the association between maternal smoking in pregnancy and obesity in the offspring. In case of a linear association the additional effects observed for paternal smoking in pregnancy might be explained by a dose effect. In case of a threshold effect with a steep increase related to small numbers of cigarettes the effect of both maternal and paternal smoking would be more likely to reflect residual confounding related to shared family environment.