Socioeconomic inequality in birth weight and gestational age in Denmark 1996–2007: Using a family-based approach to explore alternative explanations
Highlights
► Low socioeconomic position is known to be associated with increased risk of preterm birth and decreased birth weight. ► We explore the nature of this relationship by contrasting ordinary analyses with analyses where we compare children who are siblings and children who are first cousins. ► Our findings suggest that a major part of the association between maternal education and the outcomes may be explained by factors shared by mothers who are siblings.
Introduction
Children whose parents have short educations or low incomes are at increased risk of being born preterm or with a low birth weight. This association has been observed in numerous populations (Kramer, Seguin, Lydon, & Goulet, 2000; Spencer, 2003). This has also been observed in the Nordic welfare states, were educational inequalities in birth outcomes are persisting and have perhaps even grown stronger over recent years (Mortensen et al., 2008; Petersen et al., 2009).
It is not well understood why there is an association between parental socioeconomic position and pregnancy outcomes. Several studies have sought to explain the association by examining whether a set of explanatory variables can account for the association between parental socioeconomic position and health in early life (Dowd, 2007; Hafeman & Schwartz, 2009; Jansen et al., 2009; Reime et al., 2006). For example, two recent cohort studies have suggested that factors like maternal anthropometrics and smoking account for a major part of association between maternal education and birth weight (Mortensen, Diderichsen, Smith, & Andersen, 2009; Silva et al., 2010). In these studies the explanatory variables are not described as confounders, but as mediators, i.e. variables that act as intermediaries on the causal pathway from parental socioeconomic position to birth outcomes (Rothman, Greenland, & Lash, 2008). The use of the concept of mediation implicitly involves a causal claim, namely that socioeconomic position causally affects the level of the intermediary factors. But this assertion is potentially problematic. For instance, getting a college degree is something that takes place in early adulthood, whereas smoking initiation and the establishment of growth trajectories occur earlier in life. More often than not the putative intermediary variables are measured at the same point in time as socioeconomic position, which means that there is nothing that by design ensures that one precedes the other (Adair, 2008; Kraemer, Wilson, Fairburn, & Agras, 2002; National Cancer Institute, 2011; Nonnemaker, Morgan-Lopez, Pais, & Finkelstein, 2009). This does not invalidate mediational analyses, but it highlights the need for careful consideration of the alternative explanation that there may be confounding influences from antecedents of educational attainment and birth outcomes. For example, educational attainment is known to be strongly associated with parental educational attainment, even in the Nordic welfare states where equality in educational opportunities have been heavily promoted through policy (Jonsson & Erikson, 2007). It may be that early life circumstances of relevance to later-life reproductive outcomes vary according to the socioeconomic characteristics of the parental household, which may explain the educational differences in birth outcomes. One approach to this problem of confounding from early life factors would be to collect information on relevant early life circumstances, but this complicates the analyses because information on these factors may not be available or may be measured with considerable error.
Following a seminal paper by Krieger et al. the influence of early life circumstances on the relationship between socioeconomic position and health in adulthood has been studied in a number of studies of twins and non-twinned siblings (Krieger, Chen, Coull, & Selby, 2005; Madsen, Andersen, Christensen, Andersen, & Osler, 2010; Søndergaard et al., 2012). Studies of monozygotic twins have shown that differences in health according to education exists even among individuals who are genetically identical and partly/fully share early life social and material circumstances. The attraction of sibling comparisons is that when siblings that are discordantly exposed (i.e. have different socioeconomic positions) are compared, all factors shared between them are matched out. In life course epidemiology, which takes it as a premise that early life circumstances may affect health later in life, this design has been strongly advocated (Strully & Mishra, 2009). An attractive characteristic of the sibling control design is that the factors that are matched out need not be directly measured: By design all that is shared is matched out. In social epidemiology the (non-twinned) sibling design has been used to show that a substantial proportion of the association between education and cigarette smoking in adulthood is explained by factors shared by siblings (Gilman et al., 2008). It has also been used to show that increased risk of low birth weight and preterm birth among young mother is explained by factors shared by siblings, suggesting that the excess risk is not a ‘biological’ consequence of young maternal age, but is explained by the early life circumstances of the pregnant women (Lawlor, Mortensen, & Nybo Andersen, 2011). Differing opinions on the utility of these types of family designs exist: some have argued that siblings generally are as (dis)similar as unrelated individuals once shared genes are accounted for (Plomin & Daniels, 2011; Smith, 2011). According to this point of view the sibling design (when the genetic contribution is negligible) is bound to give similar results as a study of unrelated individuals, but at the cost of having to collect family data, strongly reduced statistical power and potentially other added assumptions (Kaufman, 2008). Others have argued that while this may be true siblings designs still provide the opportunity to actually test the influence of shared early life circumstance (Strully & Mishra, 2009).
The sibling design allows for control for shared early life circumstances, which may help explain the association between socioeconomic position and birth outcomes, but if the shared component is limited the results may remain confounded. Another approach would be to examine changes in socioeconomic factors in relation to changes in birth outcomes between pregnancies. This is possible because pregnancy is something that most often happens more than once to each woman. Such an analysis can be considered be an instance of the case crossover design, where the exposure for a given person at a time where an event occurred is compared with the exposure for the same person at a different time (Maclure & Mittleman, 2000). The strength of this design is that all characteristics of an individual that do not vary over pregnancies are matched out. The two family-based designs used here have been promoted as methods to improve causal inference from observational data (McGue, Osler, & Christensen, 2010; Rutter, 2007), but it is important to keep in mind that the case crossover and sibling control designs are observational designs that will only allow the analyst to identify causal effects under special circumstances. Our study is based on relatively crude analyses of whether the associations between parental socioeconomic position and birth outcomes persist when the children compared are siblings or have mothers who are siblings. The type of explanations that can this study can offer is from the onset limited to either factors shared between children who are siblings or factors shared between children whose mothers are siblings. This present study is not genetically informative and no inference on the role of genes can be drawn from this study.
The objective of this study is to examine the association between parental socioeconomic position, preterm birth and small for gestational age in three different designs: cohort, sibling control, and case-crossover. The aim is not to show that socioeconomic inequality in birth outcomes do not exist because it does. The aim is to increase the understanding of the origins of the association.
Section snippets
Population
From the Danish Medical Birth Registry we selected all live born children born in Denmark 1996–2007 (n = 794,239). The selection of the data is illustrated in Fig. 1. We excluded multiple pregnancies (n = 37,140) and stillbirths (n = 2281). Children whose grandmothers were born outside Denmark were excluded because their identity was not very likely to be recorded in Danish registries, which meant that family relations could not be reconstructed (n = 92,691). The population with linked data on
Results
In order to eligible to the study, registry linkage with the maternal grandmother of the child at risk had to be obtained. The linked and unlinked children were similar to each other in most respects (see Webtable 1). The characteristics of the children included in the study and those excluded because they did not have a sibling or a cousin in the sample are shown in Table 1.
There were only very minor differences in characteristics between those included and those excluded. The children
Discussion
Income was not associated with preterm birth or small for gestational age in Denmark around the turn of the millennium. This is in contrast to the known associations between income and health later in life in Denmark (Hoffmann, 2011). Educational differences in the risk of preterm birth and SGA were not observed in the case crossover design. The association between maternal education and SGA was the only association that was present when comparing mothers who were siblings. The results of this
Conclusions
This study suggests that a substantial part of the association between parental education, preterm birth and SGA is explained by factors that were shared by factors shared by mothers who are siblings. There was little indication that changes in socioeconomic position between pregnancies was associated with preterm birth and SGA. These findings suggest that in this sample of Danish women early life course social and biological circumstances are important in understanding the educational gradient
Acknowledgment
This study was funded by an individual post doc grant from the Danish Council for Independent Research|Medical Sciences (grant identifier number: 09-066343) and the LifeSpan Network of Excellence, funded by the European Union 6th Framework Programme. The author would like to thank the International Study Group on Income Gradients, and seminar participants at the Section of Social Medicine, University of Copenhagen, and at the Institute of Medical Epidemiology and Biostatistics, Karolinska
References (47)
Child and adolescent obesity: epidemiology and developmental perspectives
Physiology and Behavior
(2008)- et al.
A United States national reference for fetal growth
Obstetrics & Gynecology
(1996) Early childhood origins of the income/health gradient: the role of maternal health behaviors
Social Science & Medicine
(2007)Socioeconomic inequalities in old-age mortality: a comparison of Denmark and the USA
Social Science & Medicine
(2011)- et al.
The role of mediating factors in the association between social deprivation and low birth weight in Germany
Social Science & Medicine
(2006) - et al.
Danish registers on personal income and transfer payments
Scandinavian Journal of Public Health
(2011) - et al.
Study of environmental, social, and paternal factors in preterm delivery using sibs and half sibs. A population-based study in Denmark
Journal of Epidemiology & Community Health
(1999) - et al.
Change in social status and risk of low birth weight in Denmark: population based cohort study
British Medical Journal
(1997) - et al.
Parental schooling and child development: Learning from twin parent
(2009) The epidemiology of smoking during pregnancy: smoking prevalence, maternal characteristics, and pregnancy outcomes
Nicotine & Tobacco Research
(2004)