Invited Commentary: Beyond Barker—Mothers Are the Ones at Risk

Abstract

Dr. David Barker hypothesized that low birth weight (LBW) is the result of inadequate fetal nutrition, leading to increased risk of cardiovascular disease (CVD) in the offspring. This hypothesis has stimulated thousands of reports on low birth weight (LBW) and CVD risk. One problem with this association is that many LBW infants are small because they are preterm, not growth-restricted. A second problem is that maternal CVD risk factors confound the association. In an accompanying article, Lu et al. (Am J Epidemiol. 2023;192(6):866–877) address both concerns. Using population data from Sweden and Denmark, the authors estimated CVD incidence among offspring born small for gestational age (SGA). The smallest 3% had a CVD hazard ratio of 1.44 (95% confidence interval: 1.38, 1.51). Even this moderate risk mostly evaporated in sibship analysis, which controlled for unmeasured maternal CVD risk factors (hazard ratio = 1.11, 95% confidence interval: 0.99, 1.25). The risk highlighted by Barker is negligible, especially when compared with a more urgent health issue—cardiovascular risk in women with pregnancy complications. Mothers of SGA infants have up to a 3-fold CVD risk, and mothers with preeclampsia and preterm delivery have up to a 9-fold risk. Pregnancy complications thus provide an early marker of a woman’s propensity to develop CVD, and perhaps an opportunity for early intervention. From a public health perspective, Barker’s hypothesis about CVD risk in LBW offspring is less compelling than the question of CVD risk among mothers with pregnancy complications.

This article is part of a Special Collection on ABC.

This article is linked to 'Birth Weight, Gestational Age, and Risk of Cardiovascular Disease in Early Adulthood: Influence of Familial Factors' (https://doi.org/10.1093/aje/kwac223).

Abbreviations

CVD

cardiovascular disease

DAG

directed acyclic graph

LBW

low birth weight

SGA

small for gestational age

Editor’s note: The opinions expressed in this article are those of the author and do not necessarily reflect the views of the American Journal of Epidemiology.

In a series of papers starting in the 1980s, Dr. David Barker brought wide attention to the increased risk of cardiovascular disease (CVD) among persons with low birth weight (LBW). Barker proposed that undernutrition in utero “permanently changes the body’s structure, function and metabolism” (1, p. 412), leading to coronary heart disease in later life. A corollary of this causal hypothesis is that interventions designed to improve fetal growth would improve the offspring’s cardiovascular health (2). A directed acyclic graph (DAG) representing this hypothesis is shown in Figure 1.

Figure 1.

Directed acyclic graph expressing Barker’s hypothesis (1) that fetal undernutrition leads to permanent changes in the fetus that increase subsequent risk of cardiovascular disease (CVD).

There are 2 long-standing concerns about this hypothesis. First, studies supporting the association between size at birth and CVD have based their analyses largely on measures of LBW (3). LBW includes babies who are not growth-restricted but rather are small due to short gestation (4). A focus on fetal growth requires careful consideration of gestational length—for example, through analysis of infants who are born small for gestational age (SGA). Analyses of offspring risk of CVD based on SGA birth are surprisingly scarce. The second concern is the possibility of confounding by the mother’s own profile of cardiovascular risk (5). Mothers and their offspring share important CVD risk factors, including genetic susceptibility and social and behavioral characteristics. To the extent that maternal CVD risk factors reduce fetal size, they also contribute to the link between LBW and offspring CVD risk.

In this issue of the Journal, Lu et al. (6) bring valuable new data to bear on both of these concerns. The authors combined data from the national health registries of 2 Nordic countries (Sweden and Denmark) to construct a powerful analysis of offspring birth characteristics and subsequent cardiovascular morbidity. They followed 3.4 million infants into early adulthood and assessed their risk of incident CVD morbidity in relation to birth weight and gestational age. While the length of follow-up was constrained by the limited amount of time elapsed since the creation of these national registries, the large sample size provided compensatory power. The authors found that the smallest 3% of babies by gestational age had a CVD hazard ratio of 1.44 (95% confidence interval: 1.38, 1.51).

More importantly, the authors addressed in 2 ways the possibility of confounding by maternal characteristics. First, they adjusted for recorded variables related to maternal CVD risk, including family history of CVD. This adjustment reduced the CVD risk of small babies slightly, to 1.38 (95% confidence interval: 1.32, 1.45) based on model 1. The authors also conducted a sibship analysis, which controlled for more subtle but possibly important maternal and environmental factors. Sibship analysis was possible because the Swedish and Danish registries link women with their complete pregnancy histories. While sibling analyses have their own methodological limitations (7), the results were nonetheless instructive. Comparing the CVD risk of SGA babies with that of their full siblings, the elevated risk with SGA birth dropped to 1.11 (95% confidence interval: 0.99, 1.25) based on model 1. Unmeasured characteristics of the mother or her environment apparently accounted for most—if not all—of the link between SGA birth and CVD.

The Nordic data cannot rule out the possibility of direct effects of fetal growth on CVD risk, but such effects would have to be marginal, and they would occur among a tiny percentage of births. From a pragmatic standpoint, this CVD risk has little public health impact. Some might raise the objection that these Scandinavian results are not relevant to developing countries, where babies are smaller and CVD risk factors are greater. For this concern to be valid, the link between SGA birth and CVD would have to be stronger in developing countries than in developed countries. This remains to be shown—and there are indirect reasons, at least, to doubt that the association is stronger in disadvantaged populations (8).

CVD is also related to SGA birth and other pregnancy complications in a quite different way: through the mother. Women who suffer pregnancy complications are at increased risk of CVD, with relative risks that far exceed those of their offspring. Research over the past decade has established increased CVD following not just SGA birth but preterm delivery, preeclampsia, gestational diabetes, placental abruption, stillbirth, and neonatal death (9). In a meta-analysis, the relative risks of CVD after experiencing 1 complication of pregnancy ranged up to 3-fold (9). These associations become even stronger when considering full pregnancy history. Women with 2 preterm births have a 4-fold risk of dying from CVD (10). Women with preterm preeclampsia and no other pregnancies are at 9-fold risk of death (11). Pregnancy complications provide far more information about a mother’s CVD risk than about the CVD risk in her offspring.

Most of these pregnancy complications are characterized by placental pathologies, including errors of placentation and placental dysfunction. Placental pathologies are in turn more common in the presence of maternal cardiovascular risk factors. Maternal diabetes and hypertension have well-known links to placental pathologies, and even subclinical CVD risk factors may be important. For example, a maternal glucose level in the upper quartile of the normal range in healthy women is associated with increased risk of preterm delivery and preeclampsia (12).

These observations have been unified by Sattar and Greer (13) into the concept of pregnancy as a cardiac “stress test.” The authors propose that pregnancy taxes the maternal system in such a way as to reveal a CVD vulnerability that is not otherwise apparent (Figure 2). Under this hypothesis, women with greater latent risk of CVD are more likely to suffer pregnancy complications and also overt CVD later in life.

Figure 2.

Pregnancy as a “stress test” uncovering latent vascular dysfunction (expressed as pregnancy complications) that emerges later in the mother’s life as overt cardiovascular disease (13).

This concept is more fully expressed as a DAG in Figure 3. The associations between pregnancy complications and subsequent CVD risk in offspring (of which Barker’s hypothesis is one example) are subject to strong confounding by maternal CVD risk factors shared with offspring. Similarly, the link between pregnancy complications and maternal CVD risk is confounded by underlying maternal CVD risk factors. (The width of the arrows in Figure 3 represents the proposed relative strength of the causal associations.)

Figure 3.

Directed acyclic graph showing proposed causal pathways leading to associations between maternal and offspring cardiovascular disease (CVD) risk. Arrow widths represent the relative strengths of the associations. Arrows with question marks represent proposed but less certain pathways.

This DAG provides a more causally rigorous context for Barker’s hypothesis, which is contained within the small upward-curving arrow at right. For SGA birth, at least, this pathway is shown by the Nordic study to be negligible. Preterm delivery may have a more substantial association. The Nordic study suggests the possibility of a direct effect by preterm delivery, with a CVD hazard ratio of 1.21 (95% confidence interval: 1.07, 1.37) in sibship analysis. This small association could represent a direct effect of preterm delivery on offspring (as the authors suggest), or perhaps an effect mediated through postdelivery medical treatments or the developmental difficulties suffered by those born preterm (14). Regardless, preterm birth presents at most a minor contribution to CVD risk.

Meanwhile, mothers with pregnancy complications are of growing interest because of their heightened risk of CVD. CVD is not only the leading cause of death in women (15), but it is less readily diagnosed in women and often undertreated (16). Pregnancy history may offer an early signal of women’s predisposition to CVD, before full emergence of the usual clinical risk factors (17). Earlier detection of risk would allow earlier and more effective steps toward cardiac health, including changes in diet, smoking, and exercise, as well as pharmacologic prevention by aspirin or statins (18). Early detection and intervention could provide a particular benefit for racial/ethnic minorities, among whom a high prevalence of cardiovascular risk factors may contribute to their persistent excess risk of pregnancy complications. It is even possible that early intervention could improve the outcomes of a woman’s future pregnancies. Given the high recurrence rate of pregnancy complications, intervention on cardiac risk factors might be especially useful in reducing a woman’s risk of repeat complications.

Recent policy statements by cardiology societies recommend that “pregnancy history should be an integral part of cardiovascular risk assessment” (15, p. 973), while acknowledging that better data are needed (18). There are great opportunities here for collaborations between CVD epidemiologists and perinatal epidemiologists. Studies are needed to explore patterns between specific pregnancy complications and their associations with types of CVD risk. There is also a need to quantify the amount that pregnancy history might add to conventional measures of CVD risk. A related question is the extent to which subclinical CVD risk factors might be predictive of pregnancy complications.

Direct effects of pregnancy complications on maternal CVD risk are represented in Figure 3 by the downward-curving arrow on the right. The strong potential for confounding by maternal CVD risk factors requires exceptionally careful studies to establish (or rule out) such effects. Regardless of its causal underpinnings, pregnancy history remains a potentially useful marker of women’s CVD risk.

The critique of Barker’s hypothesis implied by the Nordic study does not discount the contributions of Barker’s hypothesis, particularly its promotion of life-course epidemiology. Barker’s ideas have been influential in integrating health studies across the stages of human development. That said, Barker’s hypothesis falters in fundamental ways. It focuses on just one type of pregnancy complication (small infant size). The hypothesis has been used to justify nutritional interventions that have little or no effect on birth weight, and without proven benefit to cardiovascular health (19). The Nordic data add a crucial final critique, namely that the CVD risk of small babies is too minor to be relevant. These results are discouraging for future studies of fetal growth and CVD, but there is a good alternative. The connections between pregnancy outcomes and maternal cardiovascular health present a rich opportunity for research.