Introduction
Body Mass Index (BMI) is an important marker of overall health and is strongly associated with the risk of – and mortality from – chronic diseases [
1]. In the last three decades, the prevalence of obesity has dramatically increased globally [
2], and the burden of disease attributable to high BMI has more than doubled, although this trend differs across countries [
3]. Within countries, overweight and obesity are not randomly distributed across the population, but tend to be more frequent among people with a lower educational attainment (EA) [
4] and in low socio-economic strata [
5].
One possible explanation for the association between EA and obesity is that a higher EA leads to better occupations, higher income, and higher socioeconomic status (SES). In fact, many studies use EA – as well as occupation and income – either as proxy for, or a component of, SES measures. SES is inversely related to obesity risk in global north countries [
6‐
9]. Higher education and socioeconomic status result in a better health knowledge, and provide the individual with resources that afford healthy food options [
10] and free time to engage in physical exercise [
11], i.e., behaviors associated with lower obesity risk. In addition, early research showed a residual link between obesity and cognitive deficiency, even after controlling for parental social class [
12]. Conversely, obesity itself can influence EA [
13], employment [
14], future income [
15], and other SES measures [
16,
17].
EA and BMI are both subject to intergenerational transmission. The average correlation between parent and offspring EA in Western Europe and USA is 0.39 [
18]. As for BMI, a meta-analysis of family studies on BMI transmission found a mean parent–offspring correlation of 0.19 [
19], while another meta-analysis showed that having a single parent with obesity is strongly associated with childhood obesity (average odds ratio of 3.49)[
20].
In addition to the effect of parental BMI on offspring BMI, studies have shown significant associations between parental EA and offspring BMI [
21‐
23]. Of these studies, only one included parental BMI as a covariate, with the finding that only maternal education had a significant effect [
21]. Furthermore, many studies have shown significant associations between parental SES and offspring BMI, yet none of these studies controlled for parental BMI [
24‐
30]. When including parental BMI, it is important to account for spousal correlation, or non-random mating. When spouse choice is not random with regard to a particular phenotype, but reflects resemblance between spouses that is higher that would be expected by chance, then the correlation between the phenotype of fathers and mothers needs to be modelled to avoid confounding. Overlooking spousal correlation where one exists can result in overestimating the association between one parent and offspring.
As we have laid out, BMI and EA are intertwined through various processes, and both are subject to intergenerational transmission. Prior studies have generally suggested that parental EA has a distinctive effect on the BMI of offspring, with similar findings concerning parental SES and offspring BMI. We note that these studies did not account for the correlation between parental EA and parental BMI on one hand, and parental BMI and offspring BMI on the other hand. Therefore, we conjecture that these findings—broadly stated as children growing up in low EA households being more susceptible to obesity—may be confounded by parental BMI. Simply put, it is possible that the observed association is due to the fact that EA and BMI are correlated within an individual, and the transmission of both these traits from one generation to another might give the appearance of an association between one trait (EA) in the parent generation with the other (BMI) in the offspring generation. Therefore, we utilized structural equation modeling (SEM) to examine these associations while correcting for the suspected confounding. SEM is a powerful multivariate statistical technique that allows us to test both direct and indirect effects of hypothesized causal relationships among variables [
31]. Its utility to our research question lies in its ability to account for the correlation between two causal variables in a regression analysis. The SEM approach also allows us to estimate the residual correlations between EA and BMI in the offspring generation, so that we can assess the residual association while controlling for parental transmission. This examination is important to assess the extent to which EA and BMI associations in the children are independent of the parental effects.
Discussion
Offspring BMI was negatively correlated with parental EA (
r ~ -0.07) but this correlation was low and insignificant in the full model, when accounting for parental BMI. This suggests that the effect of parental EA on offspring BMI is mainly mediated through parental BMI. Similar trends were seen when examining the influence of parental factors on offspring EA. The significant lowering of parental EA/BMI regression coefficients, when controlling for the other parental traits, supports earlier findings of shared factors influencing EA and BMI. This finding calls into question the general consensus that parental EA – and subsequently parental SES – has a direct effect on the odds of developing obesity in the offspring generation. Rather, it is an individual’s own EA that has a higher association with their obesity risk. Potential mechanism that explain these associations span a variety of social, behavioral, metabolic, and neurocognitive processes. Individuals with higher BMI are more likely to have lower self-esteem and to experience social marginalization [
34,
35]. Behavioral factors such as self-control and delayed gratification are associated with EA [
36‐
38] as well as obesity [
39,
40]. Decreased cognitive function is associated with impaired metabolic pathways associated with obesity such as insulin signaling [
41] and leptin regulation [
42]. Finally, obesity appears to have significant genetic overlap with brain and cognitive measures [
43] as well as EA [
44].
To our knowledge, this is the first study in an adult population to examine the effects of parental EA on adult offspring BMI, while controlling for parental BMI. The average age of offspring in our sample is 32 years, which means that most will have left the parental home around 10 years earlier, as the average age at which offspring leave home in the Netherlands is 22.7 years for daughters and 24.2 for sons [
45]. Still, we observe parental BMI to be correlated with their adult offspring BMI (
r = 0.18). This is in line with a literature review of studies examining parent–offspring BMI associations [
19]. Age adjusted parent–offspring BMI correlations were lower after accounting for parental EA, which shows that a portion of intergenerational BMI transmission is due to factors related to parental EA.
Within-person EA-BMI correlations were small in the offspring generation before accounting for parental factors (r = -0.15 for sons and -0.17 for daughters), with similar level for mothers (-0.15) and slightly lower for fathers (-0.10). In the offspring generation, these correlations were -0.10 after controlling for parental transmission. This persistence of association suggests that the relationship between EA and BMI is largely independent of parental factors, which implies that interventions aimed at improving EA (and consequently SES) can translate into desirable changes in BMI, irrespective of parental EA and BMI.
Gender did not play a significant moderating role in our model. Transmission coefficients were largely similar from fathers and mothers to sons and daughters, with the exception of BMI transmission from mothers to sons, which was lower than other parent–offspring combinations, but had the same direction of association (i.e. positive). The absence of gender differences confirms findings of prior studies [
20], although few studies have reported differences between fathers and mothers in BMI transmission [
46,
47].
Parents in our sample exhibited moderate levels of spousal correlation for BMI. For EA, the association was high. The observed spousal correlation for BMI in our study (
r = 0.23) is somewhat higher than that reported in most other studies, averaging at 0.15 [
48]. Increased rates of spousal correlation over birth cohorts have been hypothesized to have contributed to the rise in obesity prevalence [
49]. Indeed, the odds of offspring obesity increase markedly when both parents have obesity [
50]. From our study design, it is unclear whether this correlation existed prior to marriage/cohabitation (due to phenotypic assortment or social homogamy) or developed with time (due to marital interaction). However, the latter scenario would involve an increase over time in spousal correlation for BMI, which is generally not found [
48].
The main strengths of this study are a large sample size and age range for parents and offspring, as well as use of multiple offspring within families. Our study sample covers different geographic areas and socioeconomic classes in the Netherlands. Recruitment of twins – considered representative of the general population [
51] – and their families into the NTR was done through multiple channels including city council registries, leading to a sample that is reasonably representative of the Dutch population. Although the study relied on self-reported measures for height, weight, and EA for parents and offspring, the reliability in self-reporting of height and weight is good: in an analysis of a subsample of 6,026 individuals, we observed a correlation of 0.93 between self-reported BMI and BMI measured by a research nurse or assistant [
52]. We analyzed BMI as a continuous variable rather than a (clinical) binary variable (obese/overweight vs. normal weight), because we wanted to address the process of transmission as it pertains to the full range of BMI in the general population in the Netherlands. In this approach we assume that the process of transmission as we identify it is relevant to the full range of BMI and EA. We note that it is possible to fit our present model (see Fig.
1) to a binary BMI variable, but this would merely result in a loss of information. Specifically, the analysis in the case of a binary BMI variable is based on the liability threshold model, in which the model is fitted to latent continuous BMI variables. There is no advantage to this compared to fitting the directly to observed continuous BMI variables, but there is the disadvantage of a loss of power associated with loss of information. For this reason, we opted to use BMI as a continuous variable in our analysis.
There are also some limitations to the current study. While BMI is a common, convenient measure of obesity, other methods, such as skin fold thickness and percent body fat from dual energy X-ray absorptiometry, may provide more accurate estimates. In our analyses, we have modeled the process of transmission of EA and BMI from adult parents to adult offspring. Our resulting model is supposed to represents this process as it takes place in the general Dutch population, and with respect to the full range of BMI. As such, this model is based on the assumption that the process is the same regardless of the actual level of BMI in the parents or their offspring. We recognize that it is possible that the transmission process may differ (e.g., in terms of the parameters) in extremes of the BMI distribution, i.e., in the obese and underweight subpopulations. Detecting such heterogeneity is a statistically challenging task, which is beyond our present aim.
Our results pertain to adult offspring who generally have left the parental household. A next step in future research is to examine whether the associations and transmission results based on adult offspring are also seen in younger offspring who still live with their parents, and likely share more of the home environment. This will facilitate the disentangling of the different aspects by which parents influence their offspring’s probability of having overweight or obesity. This is important to inform policy and interventions aimed at reducing the prevalence of these conditions. Our results suggest that improving the SES of a household may not on its own be sufficient to reduce the odds of obesity among offspring. It is imperative that policies focus on long term strategies such as educational expansion and improving social mobility, which would have a more pronounced impact on obesity rates one generation after another.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.