In this issue of Paediatric and Perinatal Epidemiology, Segovia Chacón and her associates¹ used linked data on 1,100,000 live births from the Swedish Birth Defects and Health Registries to ask whether estimates of the prevalence of major congenital anomalies differ by age of the child, a question that colours how we interpret birth defect rates observed in exposure cohort studies. The authors found that the observed prevalence of all major malformations and many anatomic classes of major malformations do increase monotonically from birth to 5 years of age. However, for all major congenital anomalies and most anatomic classes of major malformations, the increase is much smaller after 1 year of age than between birth and 1 year. Based on these observations, Segovia Chacón and her colleagues make the reasonable and practical suggestion that measuring the frequency of major congenital anomalies diagnosed by 1 year of age is sufficient for most studies looking for an association between maternal drug treatment during pregnancy and infant malformations.

Finding that the prevalence of major congenital anomalies increases from birth through early childhood is, of course, no surprise. Similar observations have been made repeatedly in other pharmacoepidemiological studies for over 50 years. The current study and Nordic records linkage studies generally are distinguished from many other investigations by the consistent quality of their data and their ability to link individual records accurately. These factors permit a wide variety of outcomes to be identified on a population-wide basis for many years after birth.

The current study employs a well-established and standardised design, but this design does not reflect a contemporary understanding of developmental biology and teratogenic mechanisms. For example, the outcomes of interest are ‘major congenital malformations’ as a group, as well as anatomic classes of major malformations. The range of malformations included as ‘major’ and the grouping of anomalies into anatomic classes are inconsistent with current knowledge. ‘Major’ congenital anomalies are defined as structural changes that have significant medical, surgical, social or cosmetic consequences for the affected individual.² Thus, ‘major congenital malformations’ include not only the lethal conditions of Anencephaly (International Classification of Diseases [ICD] version 10 code Q00.0) and Craniorachischisis (ICD-10 Q00.1) but also Absence and agenesis of lacrimal apparatus (ICD-10 Q10.4), i.e., blocked tear duct. Although Absence and agenesis of lacrimal apparatus is included as a major malformation with a cumulative detection rate that increases with age in Segovia Chacón's Table 1, this anomaly is usually completely treatable by massaging the tear duct in babies under 6 months of age or by surgical probing if massage does not work.

The grouping of major malformations into anatomic classes ignores the fact that teratogenic exposures are more likely to produce a recurrent pattern of multiple minor and major congenital anomalies throughout the body than several different major malformations within a single anatomic class. It would, therefore, make more sense to group anomalies by suggested ‘syndrome’ pattern or postulated pathogenic mechanism than by anatomic class.³

The authors' concern about whether some cases of microcephaly are acquired rather than congenital illustrates the issue of anomaly classification, but the problem goes deeper than age at diagnosis. Microcephaly is sometimes not really a malformation (i.e., an alteration of a primary developmental process) but rather a disruption (i.e., the breakdown of an originally normal developmental process),³ as in the severe microcephaly that occurs with Zika virus embryopathy.⁴ In other cases, microcephaly may be a manifestation of universal growth impairment that affects height and weight to a similar degree as head circumference.

In addition, the traditional focus of birth defects registry studies on major malformations excludes not only recurrent patterns of minor abnormalities and generalised impairment of growth but also neurodevelopmental abnormalities. All of these problems are important components of many teratogenic embryopathies. Neurodevelopmental abnormalities, such as severe intellectual disability or autism, are a particular concern for many pregnant persons who are worried about the possible adverse effects of an exposure on their embryo or foetus. Nordic linked-records systems provide an exceptional opportunity to study the effects of maternal gestational exposures on neurobehavioral outcomes in the children,^5-7 but neurodevelopmental abnormalities are rarely included with congenital anomalies in studies of possible teratogenic effects of maternal pregnancy exposures.

Another area in which the standard methodology of birth defects registry studies is out of date relates to how chromosomal abnormalities and genetic diseases are considered. Segovia Chacón et al. found that the proportion of children with chromosomal or other genetic anomalies ranged from 0.16% at birth to 0.32% at 3 years of age. The use of exome and genome sequencing has dramatically increased the number of patients in whom a genetic cause can be recognised, and it seems likely that the proportion of children with recognised chromosomal or other genetic abnormalities is much greater today than it was when data collection for this study began. Even at 0.32%, the proportion of genetic disorders recognised in the Swedish Medical Birth Register is an order of magnitude less than the estimated prevalence of individuals with genetically caused rare diseases.⁸

Segovia Chacón's study excluded children with chromosomal or other genetic anomalies from the group with major malformations unless the child also had a major structural anomaly. The reasoning behind this exclusion was that children with genetic disorders have a higher prevalence of ‘nongenetic’ malformations, and maternal medication use is not likely to cause genetic changes in the embryo/foetus. It certainly is true that constitutional genetic mutations cannot be caused by a teratogenic exposure that occurs after implantation, but we now know that most major malformations have a complex origin, with both genetic and nongenetic (e.g., teratogenic) predisposing factors. Embryos with a genetic predisposition to a malformation may, therefore, act as ‘canaries in the coal mine’, being at greatly increased risk of developing a malformation after suffering a teratogenic exposure in comparison with embryos without a genetic predisposition. By analysing cases with a genetic condition such as a chromosomal abnormality or Mendelian syndrome as a separate group, rather than excluding them from a study, it may be possible to detect a signal of teratogenic risk with fewer total exposures.

Nordic record-linkage studies have been a cornerstone of pharmacoepidemiology for many years. Updating the design of these studies in consideration of our contemporary understanding of how teratogenic effects occur and can be recognised in humans could make these studies more valuable.

在本期《儿科和围产期流行病学》中，Segovia Chacón 和她的同事¹使用瑞典出生缺陷和健康登记处 1,100,000 名活产儿的关联数据来询问主要先天性异常患病率的估计是否因儿童年龄而异，这是一个问题这影响了我们如何解释暴露队列研究中观察到的出生缺陷率。作者发现，观察到的所有主要畸形和许多解剖学类别的主要畸形的患病率确实从出生到 5 岁单调增加。然而，对于所有主要先天性异常和大多数解剖学类别的主要畸形，1 岁后的增加比出生至 1 岁之间的增加要小得多。基于这些观察，Segovia Chacón 和她的同事提出了合理且实用的建议，即测量 1 岁时诊断出的主要先天性异常的频率足以满足大多数寻找孕期药物治疗与婴儿畸形之间关系的研究的需要。

当然，发现从出生到幼儿期主要先天性异常的患病率不断增加，这一点并不令人意外。50 多年来，其他药物流行病学研究中也多次出现类似的观察结果。当前的研究和北欧记录链接研究通常与许多其他调查不同，其数据质量一致，并且能够准确链接各个记录。这些因素使得在出生后多年内能够在全人群范围内确定各种结果。

目前的研究采用了完善的标准化设计，但这种设计并没有反映当代对发育生物学和致畸机制的理解。例如，感兴趣的结果是“主要先天畸形”作为一个群体，以及主要畸形的解剖学类别。被列为“严重”的畸形范围以及将异常按解剖类别分组与当前知识不一致。“重大”先天性异常被定义为对受影响个体产生重大医疗、外科、社会或美容后果的结构变化。²因此，“重大先天畸形”不仅包括无脑畸形（国际疾病分类 [ICD] 第 10 版代码 Q00.0）和颅裂（ICD-10 Q00.1）等致命病症，还包括泪腺缺失和发育不全（ ICD-10 Q10.4)，即泪管阻塞。尽管在 Segovia Chacón 的表 1 中泪道缺失和发育不全被列为主要畸形，累积检出率随着年龄的增长而增加，但这种异常通常可以通过按摩 6 个月以下婴儿的泪管或手术探查来完全治疗。如果按摩不起作用。

将主要畸形按解剖类别分组忽略了这样一个事实：与单个解剖类别中的几种不同主要畸形相比，致畸暴露更有可能在全身产生多种轻微和主要先天性异常的复发模式。因此，通过建议的“综合症”模式或假设的致病机制对异常进行分组比通过解剖类别更有意义。³

作者对某些小头畸形病例是否是后天性而非先天性的担忧说明了异常分类的问题，但问题比诊断时的年龄更深。小头畸形有时并不是真正的畸形（即初级发育过程的改变），而是一种破坏（即最初正常发育过程的破坏）³，如寨卡病毒胚胎病引起的严重小头畸形。⁴在其他情况下，小头畸形可能是普遍生长障碍的一种表现，对身高和体重的影响程度与头围相似。

此外，出生缺陷登记研究的传统重点是主要畸形，不仅排除了反复出现的轻微异常和普遍的生长障碍，还排除了神经发育异常。所有这些问题都是许多致畸胚胎病的重要组成部分。神经发育异常，例如严重智力障碍或自闭症，是许多孕妇特别关注的问题，他们担心接触这种物质可能会对胚胎或胎儿产生不利影响。北欧链接记录系统提供了一个特殊的机会来研究孕产妇妊娠暴露对儿童神经行为结果的影响，^5-7，但在孕产妇妊娠暴露可能致畸影响的研究中，神经发育异常很少被纳入先天性异常。

出生缺陷登记研究的标准方法已经过时的另一个领域涉及如何考虑染色体异常和遗传疾病。塞戈维亚查孔等人。研究发现，患有染色体或其他遗传异常的儿童比例从出生时的0.16%到3岁时的0.32%不等。外显子组和基因组测序的使用极大地增加了可以识别遗传原因的患者数量，而且今天被识别出染色体或其他遗传异常的儿童比例似乎比收集数据时要大得多。这项研究开始了。即使是 0.32%，瑞典医学出生登记所承认的遗传性疾病比例也比遗传性罕见疾病个体的估计患病率低一个数量级。⁸

塞戈维亚·查孔（Segovia Chacón）的研究将患有染色体或其他遗传异常的儿童排除在严重畸形组之外，除非该儿童还存在严重的结构异常。这种排除背后的原因是，患有遗传性疾病的儿童“非遗传性”畸形的患病率较高，并且母亲的药物使用不太可能导致胚胎/胎儿的遗传变化。诚然，体质性基因突变不能由植入后发生的致畸暴露引起，但我们现在知道，大多数主要畸形具有复杂的起源，具有遗传和非遗传（例如致畸）诱发因素。因此，具有畸形遗传倾向的胚胎可能就像“煤矿里的金丝雀”，与没有遗传倾向的胚胎相比，在遭受致畸暴露后发生畸形的风险大大增加。通过将患有染色体异常或孟德尔综合症等遗传性疾病的病例作为一个单独的组进行分析，而不是将它们排除在研究之外，也许可以用较少的总暴露量来检测致畸风险的信号。

北欧记录关联研究多年来一直是药物流行病学的基石。考虑到我们当代对致畸效应如何发生以及如何在人类中识别的理解，更新这些研究的设计可以使这些研究更有价值。