Minorities’ Diminish Returns of Parental Education in Reducing Childhood Body Mass Index

Background: Considerable research has documented the effects of race and socioeconomic status (SES) on childhood body mass index (BMI) and obesity. However, less is known about the intersectional effects of race and family SES on childhood BMI. Purpose: This study tested racial by SES variation in BMI among American 9-10 years old children. Built on Minorities’ Diminished Returns (MDRs), we expected a weaker family SES effect on childhood BMI for non-White than White and children. Methods: For this cross-sectional study, data came from the Children Brain Cognitive Development (ABCD) study, a national multi-center investigation of child development in the US. This study included 18441 BMI observations, 9-11-year-old children. The independent variables were family SES (parental education). Moderator was race. The primary outcome was BMI. Age, sex, ethnicity, and parental marital status were the covariates. To analyze the data, we used mixed-effect regression models. Results: High parental education and race were associated with BMI. We found an interaction between race and parental education with non-White children with highly educated parents still having a high BMI. Conclusions: For American children, BMI is shaped by the intersection of race, gender, and family SES. Children from highly educated families remain at risk of high BMI. Disparities in BMI and obesity should be approached through an intersectionality lens.

The societal and structural causes of these MDRs are partially known. It has been known that high SES racial minority families experience high levels of discrimination .
There is also a growing literature on MDRs of family SES on childhood obesity and BMI. A 15 yearfollow-up study used the FFCWS data; high family SES at birth reduced BMI of White but not Black children at age 15 (Assari, Thomas, et al., 2018). In another study, a high family income better reduced the BMI of White than Black children (Assari, 2018d). In another study using Health and Retirement Study (HRS), education better reduced BMI for White than Black people (Assari, Nikahd, Malekahmadi, Lankarani, & Zamanian, 2016). Although these findings suggest that SES better reduces the BMI of Whites than Blacks across age groups, there is limited knowledge on whether this pattern applies to all racial groups, or it is only specific to Black people.
To extend the existing knowledge on the MDRs and complexities regarding social determinants of childhood BMI in the US, this study explored race by family SES variation in BMI among 9-10 years old American children. We expected race by SES differences in BMI due to weaker SES effects on BMI for non-White than White children (Assari, 2017c;Assari, 2018;.

Design and Settings
We conducted a secondary analysis of the ABCD study data (Alcohol Research: Current Reviews Editorial, 2018; Casey et al., 2018;Karcher, O'Brien, Kandala, & Barch, 2019;Lisdahl et al., 2018;Luciana et al., 2018). With a cross-sectional design, we applied data from wave 1 of the ABCD study.
ABCD is a national, state-of-the-art brain imaging study of childhood brain development (Alcohol Research: Current Reviews Editorial, 2018; Auchter et al., 2018). The ABCD study's advantages include a national sample, a large sample size, a large sample of Blacks and Latinos, available data, robust brain development measures, and considerable socioeconomic factors (Alcohol Research: Current Reviews Editorial, 2018; Casey et al., 2018;Karcher et al., 2019;Lisdahl et al., 2018;Luciana et al., 2018).

Sample and Sampling
Our analysis unit was 14881 BMIs that belonged to 11,000+ children who had participated in the ABCD study. BMIs were nested to individuals who were nested to families nested to 21 sites (across states).
This ABCD sample was primarily recruited through the school systems with sampling (school selection) informed by race, ethnicity, sex, SES, and urbanicity. More details of ABCD sampling are published elsewhere (Garavan et al., 2018). Eligibility criteria for this study were complete data for our study variables, regardless of race, ethnicity, or sex. All participants were children between the ages of 9 and 11 and had valid data on BMI, race, SES, and our confounders.

Study Variables.
The study variables included race, age, gender, family SES (parental education), parental marital status, and BMI.

Primary Outcome
The primary outcome was BMI, measured rather than self-reported. Children's BMI was measured based on participants' height and weight. Height and weight were measured in feet/inches and pounds, respectively. Measures occurred between two or three times. The ABCD data has already calculated a BMI measure in the DEAP data set. After converting height and weight to meters and kilograms, BMI is calculated by dividing weight (kilograms) by height squared (meters squared). In this study, each individual could have one or two BMI observations. If present, the 2 nd BMI measurement one was one year later than the baseline measurement. Given the observations' nested nature, we ran models that allow adjustment for the data's non-independence (nested) nature.

Moderator
Race. Race, self-identified, was a nominal variable: Black, Asian, Other/Mixed, and White (reference).

Independent Variable
Parental educational attainment. Parents reported their years of schooling. This variable was operationalized as a categorical variable: Less than high school diploma (reference), high school diploma, some college, college graduation, graduate-level education.

Confounders
Age. Age was a continuous measure in months. Parents reported the age of the children.
Sex. Sex was 1 for males and 0 for females.
Parental marital status. Parental marital status was 1 for married and 0 for any other condition (reference).
Ethnicity. Parents were asked if they are of Latino ethnic background. This variable was coded as = 1 for Latino and 0 for non-Latino.

Data Analysis
We used the Data Analysis and Exploration Portal (DEAP) for data analysis. Frequencies (n and %) and mean [standard deviations (SDs)] were reported to describe the variables in the pooled sample and by race. We used the ANOVA test and Chi-square in the pooled sample to estimate bivariate analyses between the study variables. To perform our multivariable analyses, we performed mixed-effect regression models. Figure 1 shows the distribution of our independent variable, dependent variable, and the assumption test for our regression model. The independent variable was the family SES (parental education). The outcome was BMI. All these models controlled for ethnicity, age, sex, and parental marital status. All models were performed in the pooled sample. Model 1 did not have interaction effects.
Model 2 was performed with interaction terms between parental education and race. Unstandardized

Ethical Aspect
Our analysis was exempt from a full review. However, the ABCD study protocol was approved by the University of California, San Diego (UCSD) Institutional Review Board (IRB) (Auchter et al., 2018).

Descriptives
The sample included 14881 BMI observations from 9-11 years old children. Table 1 presents the descriptive statistics of the pooled sample and also by race. White children had the highest income and parental education, and Black children had the lowest parental education and income.    Table 2 reports the results of a pooled sample regression model. Model 1, which only included the main effects, showed that high parental education was associated with lower BMI levels ( Figure 2).  Table 3 reports the results of Model 2. This model showed that parental education and race interact, meaning that the effect of parental education on BMI was weaker for non-White than White children ( Figure 3).

Discussion
Although higher family SES (parental education) is associated with a reduced childhood BMI (1st finding), family SES shows a weaker effect for non-White and White children (2nd finding).

Conclusions
In summary, in this study with a large national sample, high family SES correlates with lower childhood BMI. However, this protective effect is weaker for non-White than White children. That means childhood BMI is shaped by the intersection of race and SES rather than their separate and independent effects. As such, eliminating the racial and economic disparities in childhood BMI and obesity requires an intersectional approach.
Author's Funding: Support received from the following NIH grants: 2U54MD007598, U54 TR001627; CA201415-02, 5S21MD000103, R25 MD007610, 4P60MD006923, and 54MD008149. implemented the study and provided data but did not necessarily participate in the analysis or writing of this report. This manuscript reflects the views of the authors and may not reflect the opinions or views of the NIH or ABCD consortium investigators. The ABCD data repository grows and changes over time.