Richards and Darrow Respond to "Methodological Research on Pregnancy Weight Gain".
Richards and Darrow Respond to "Methodological Research on Pregnancy Weight Gain".
复制标题
理查兹和达罗回应“妊娠体重增加的方法学研究”。
DOI：
10.1093/aje/kwad083
复制
发表时间：
2023
影响因子：
5
通讯作者：
Darrow,LyndseyA
中科院分区：
文献类型：
--
作者：
Richards,Megan;Darrow,LyndseyA
我们在研究中选择了这些结局（小于胎龄儿、剖宫产和低出生体重儿），因为它们在分娩时表现出不同程度的胎龄依赖性，并且可以在分娩时使用出生记录进行有效的横断面研究。事实上，早产是造成出生体重低的一个原因，这是将其列入的主要原因。通过呈现与胎龄有不同程度相关性的结果，我们证明，当结果与分娩时的胎龄相关时，不同的方法可以产生显着不同的估计。我们认为这些结果与妊娠期体重增加（GWG）的研究有关，妊娠期体重增加与分娩时胎龄是许多结局的风险因素，包括新生儿死亡率，神经发育，儿童人体测量，哮喘和呼吸系统疾病。例如，研究与早产相关的儿童哮喘的研究人员应该知道，与直接调整胎龄的模型相比，使用外部导出的z评分调整胎龄的模型的估计值可能会有意义的差异。我们承认Hutcheon和Platt对我们的声明的批评，关于在给定胎龄的持续妊娠和分娩之间GWG分布存在差异的任何时候都可能混淆的预期。z评分（基于正在进行的妊娠）和分娩时胎龄之间缺乏独立性，这也反映了GWG对早产的无偏倚因果影响。例如，在我们的研究结果中观察到，z评分较高的肥胖妇女的低出生体重风险显著较高（其他方法中不明显的相关性），这可能反映了高GWG对分娩时胎龄的因果影响，分娩时胎龄的残留混杂，或两者兼而有之。考虑到我们在特定体重指数亚组中观察到的GWG z评分相关性的大小和方向，以及Hinkle等人发表的先前模拟研究的结果。(3)，我们得出的结论是，离群z评分结果最有可能反映残留混杂，但由于我们不知道真相，我们显然不能肯定地说这一点。我们选择的所有3种方法都依赖于分娩时的累积GWG;然而，重要的是要认识到分娩时的GWG本身是整个妊娠期间GWG的功能。我们强调，即使在一个完美指定的z评分表，未测量的常见原因异常体重增加在怀孕和早产（如，母亲激素的档案，不足的母亲血容量扩张）将导致混淆。在这些情况下，调整分娩时的胎龄将完全阻断通过早产到围产期结局的所有后门路径，但从正在进行的妊娠中得出的完美指定的z评分不会。在我们看来，这种无法测量的混杂因素的可解释性已经为回答GWG是否对分娩时胎龄有因果影响的问题带来了障碍。一个令人信服的答案将大大有助于为这一领域的首选方法提供信息。
We selected the outcomes in our study (small for gestational age, cesarean delivery, and low birth weight) because they exhibited varying degrees of dependence on gestational age at delivery and could be validly studied cross-sectionally at the time of delivery using birth records. Indeed, the fact that preterm birth is a cause of low birth weight was the main reason for its inclusion. By presenting results for outcomes that have varying degrees of association with gestational age, we demonstrate that different methods can produce dramatically different estimates when the outcome is associated with gestational age at delivery. We believe these results are relevant to the study of gestational weight gain (GWG) in relation to many outcomes for which gestational age at delivery is a risk factor, including neonatal mortality, neurodevelopment, child anthropometric measures, asthma, and respiratory disease. For example, researchers examining childhood asthma, which is associated with preterm birth, should know that estimates from models using an externally derived z score to adjust for gestational age could differ meaningfully compared with estimates from models that directly adjust for gestational age. We acknowledge Hutcheon and Platt’s criticism of our statement regarding the expectation of confounding anytime the GWG distribution differs between ongoing pregnancies and births at a given gestational age. Lack of independence between z score (based on ongoing pregnancies) and gestational age at delivery could additionally reflect an unbiased causal effect of GWG on preterm birth. For example, the substantially higher risk of low birth weight observed in our results for obese women with higher z scores (an association not apparent in the other approaches) could reflect a causal effect of high GWG on gestational age at delivery, residual confounding by gestational age at delivery, or a combination of both. Given the magnitude and direction of the associations we observed with GWG z scores in specific body mass index subgroups, and given findings from a previous simulation study published by Hinkle et al.(3), we concluded that the outlier z score results were most likely to reflect residual confounding, but since we do not know the truth, we obviously cannot say this with certainty. All 3 of our chosen methods relied on cumulative GWG at delivery; however, it is important to recognize that GWG at delivery itself is a function of GWG throughout pregnancy. We emphasize that even under a perfectly specified z score chart, unmeasured common causes of abnormal weight gain during pregnancy and preterm delivery (eg, maternal hormone profiles, insufficient maternal blood volume expansion) would cause confounding. In these scenarios, adjusting for gestational age at delivery would fully block all backdoor paths through preterm delivery to a perinatal outcome of interest, but the perfectly specified z score derived from ongoing pregnancies would not. In our view, the plausibility of this kind of unmeasured confounding has presented obstacles to answering the question of whether GWG has a causal effect on gestational age at delivery. A convincing answer would go a long way toward informing preferred methods in this area.