cells rather than more metabolically relevant tissues. However, changes in white blood cell methylation have been observed in response to a number of in utero exposures. For example, a study of individuals currently in their sixties but exposed in utero to the Dutch Hunger Winter demonstrated altered methylation at a number of loci known to be involved in growth and metabolism [107]. Investigators in the Pregnancy and Childhood Epigenetics Consortium meta‐analyzed the association between pre‐pregnancy maternal BMI and methylation at over 450,000 sites in newborn blood DNA across 19 cohorts (9340 mother‐newborn pairs) [108]. In newborns, after adjustment for cell proportions, maternal BMI was associated with small methylation variation at 86 sites throughout the genome. At 72/86 sites, the direction of the association was the same in newborns and adolescents, suggesting the persistence of signals. In addition, epigenome‐wide analyses of blood from children at two different ages identified changes in DNA methylation that were associated with exposure in utero to maternal obesity and/or gestational diabetes [109]. However, changes in methylation were generally small in all analyses (<5%). Epigenome‐wide analysis has also identified changes in cord blood DNA methylation that are associated with maternal smoking during pregnancy. In one study, methylation of the locus GFI1 explained 12–19% of the reduction in birth weight resulting from maternal smoking [110]. Another study identified differences in placental methylation at seven sites that mediated the association between prenatal smoking and birth weight [111].
Studies in animal models have allowed the effects of changes in the early environment to be studied in metabolically relevant tissues. Although such studies enable causal effects of in utero exposures on the offspring epigenome to be determined, conclusions regarding the causal effects of specific epigenetic changes on the phenotypic outcome (e.g. offspring obesity) cannot be made. A study in rats demonstrated that high‐fat diet feeding during pregnancy caused hypermethylation of the promoter of POMC in the arcuate nucleus of the hypothalamus of the offspring, which was associated with increased feeding in response to a high‐fat diet [112].
Mediating metabolic factors
There has been much discussion of potential factors that could mediate the effects of suboptimal early exposures on long‐term risk of obesity. Because central regulation of energy balance plays a critical role in determination of weight regulation, development of neuronal pathways in the hypothalamus within the brain has been an area of interest. Animal studies have identified a number of hormonal and metabolic factors that influence development of hypothalamic circuits, including glucose, leptin, ghrelin, and insulin [113–116]. These factors can all be regulated by diet. Therefore, exposure to inappropriately high or low levels of these factors during critical periods of development could provide mechanisms by which nutrition during critical periods of development could impact on later obesity risk. However, translating such findings to humans is challenging.
Gut microbiota
The role of the gut microbiota in influencing obesity risk is much debated. The gut microbiome includes trillions of micro‐organisms that produce various metabolites, including short‐chain fatty acids and bile acids. Variation in microbiome composition could influence metabolism of the host and, therefore, potentially energy balance. This hypothesis is supported by studies in which the microbiomes of human twins discordant for obesity were transplanted into germ‐free mice. Those mice that obtained microbiota from the obese twin gained more fat mass than those that received the microbiota from the lean twin [117]. This finding is additionally supported by studies in mouse models in which transplantation of gut microbiota from an obese mouse donor to a germ‐free recipient results in a significant increase in body fat in the recipient [118]. Several investigators have seen higher BMI and rates of obesity among children or adolescents born by cesarean versus vaginal delivery [119]. One possible explanation for this difference is a different microbiome acquired by route of delivery, although alternatively, the effect may be explained by higher maternal BMI and child weight at birth. Evidence from animal husbandry as well as some human studies have demonstrated an association between antibiotic exposure during the first year of life and subsequent weight gain, risk of overweight, and central adiposity, which could suggest an association between the gut microbiome in early life and later obesity risk [120]. However, in large human studies, effect sizes for early‐life antibiotic exposure and later BMI are quite small [121] suggesting that such a mechanism is unlikely to be a major contributor to obesity rates in the population.
Estimating population attributable risks
The combined predictive value of these developmental risk factors for childhood obesity is substantial. Although each risk factor may be modestly associated with childhood obesity, collectively they may result in enormous differences for populations. An analysis of the Project Viva cohort found that mid‐childhood obesity prevalence at age seven was approximately 30% among children with four early life risk factors (maternal smoking and excessive gestational weight gain during pregnancy; short duration of sleep at 6 months and breastfeeding duration <12 months), compared with only 6% among those who had none of the four [122]. Results were similar in the GUSTO cohort from Singapore [123]. Another group has estimated that 47.2% (95% CI: 30.9%, 63.5%) of type 2 diabetes in youth could be attributed to intrauterine exposure to maternal diabetes and obesity [124]. In a meta‐analysis of data from 162,129 mothers and their children from 37 pregnancy and birth cohort studies from Europe, North America, and Australia, the proportions of childhood overweight/obesity prevalence attributable to maternal overweight, maternal obesity, and excessive gestational weight gain ranged from 10.2 to 21.6% [35].
Risks are likely even higher among certain subgroups of the population. In the United States, rates of obesity are especially high among children from lower‐income families as well as children who are Black, Hispanic, and other race/ethnicity compared with white or Asian‐American [125]. Racial disparities in childhood overweight appear to be largely explained by potentially modifiable early life risk and protective factors [126,127]. Interestingly, among low‐income and nutritionally at‐risk children aged 2–4 years enrolled in the Federally‐funded US Women, Infants, and Children (WIC) Program, the overall crude prevalence of obesity actually decreased from 15.9% in 2010 to 13.9% in 2016 [128]. Promising evidence suggests that changes in the WIC food package that provided healthier food purchases during pregnancy and the early postpartum years within the WIC program may be responsible for these declines [129].
Implications for policy and practice
In summary, strong evidence has identified a number of early life experiences that predict later obesity risk. However, we should be cautious about directly translating these observational associations into recommendations for public health or clinical practice, especially for exposures related to weight, weight gain, or growth. Beyond the general concern that it is difficult to fully eliminate confounding from observational studies, exposures such as “maternal obesity” or “fetal growth” do not directly map to a target trial. For some of the exposures discussed above, including maternal diet quality or smoking, one might imagine the intervention trial that would mimic with our observational analysis – for example, advice to stop smoking or not to consume sugary beverages during pregnancy. However, there are many likely paths that lead to, for example, maternal obesity entering pregnancy. Even if maternal obesity is a strong, consistent predictor of offspring obesity independent of likely confounding factors, we don’t know when and how we might intervene to interrupt this link. Should we act on the woman’s diet quality, her overall caloric intake, her physical activity, or all of these, and which would be most effective? Should the intervention occur 1 month, 1 year, or 10 years prior to conception? Because of these many complexities in a causal interpretation of weight‐related exposures, we should think of many of these factors as risk markers rather than causal risk factors.
Despite these considerations, we can use this knowledge to identify children at higher risk for subsequent obesity. We, therefore, encourage providers caring for children to take a complete health history going back to the prenatal
