Abstract
Objective:
The aim of this study was to identify longitudinal trajectories of conjoint development of executive function (EF) and obesity among a diverse sample of poor, rural youth and to evaluate individual differences in infant growth, parental BMI, and cumulative risk.
Methods:
Participants included 948 youth from the Family Life Project. Child anthropometrics were measured at 2 and 6 months and at 2, 3, 4, 5, 7, and 12 years. EF tasks were administered at 3, 4, and 5 years. Mothers reported youth birth weight, parental height and weight, and cumulative risk indicators.
Results:
Multidimensional growth mixture modeling identified three classes: “High EF – High Obesity Resilience”; “Low EF – Delayed-Onset Severe Obesity”; and “Low EF – Early-Onset Severe Obesity.” Youth in the low-EF, early-onset class displayed higher birth weight and BMI at 6 months, whereas the low-EF, delayed-onset class had rapid weight gain during infancy, parents with class II obesity, and greater cumulative risk and was more likely to be Black and female.
Conclusions:
Despite increased obesity risk among this sample, the majority of youth exhibited higher EF and some degree of obesity resilience. Youth with EF deficits displayed the greatest risk for severe obesity but had differing BMI trajectories and obesity risk profiles, which has implications for obesity intervention.
Introduction
Pediatric obesity remains a public health issue (1). Executive function (EF), a collection of potentially modifiable, integrated neurocognitive processes involved in higher-order cognition (2), has been high-lighted as a potential protective factor against obesity. EF undergoes rapid structural and functional growth between approximately 3 and 5 years of age (3) and it is necessary for self-regulated, purposeful, and goal-directed thought and behavior, including behaviors associated with obesity (4). Three commonly agreed upon EF processes are as follows: inhibitory control (the capacity to inhibit an automatic or pre-potent thought, emotion, or behavior in favor of a less desirable but more beneficial response), cognitive flexibility (the capacity to fluidly shift the focus of attention from one dimension of a given set of stimuli to a second less salient dimension), and working memory (the ability to retain multiple pieces of transitory information, “on-line,” in the mind, for potential manipulation) (5).
Low EF has been linked to unhealthy food intake, low physical activity, sedentary behavior, and obesity among children and adolescents (6,7). However, literature linking EF and obesity in general has at least two limitations. First, the vast majority of this research explores cross-sectional associations between EF and obesity in late childhood and adolescence, well after periods of rapid EF development. Establishing temporal precedence between EF and weight status is important because it suggests that EF could function as a target mechanism for obesity prevention programs, as is the case for other health behavior outcomes (8). A second limitation of the field is its overreliance on variable-centered statistical analyses to test for links between EF and obesity. Alternatively, person-centered analytic approaches identify a parsimonious set of distinct, homogeneous subpopulations of participants based on observed indicators of EF and obesity. A limited number of cross-sectional, person-centered approaches have been used to examine subgroups of participants based on weight loss strategies and parenting characteristics associated with children’s BMI (9). For example, Huh and colleagues (10) used latent class analysis, a person-centered analytical approach, to identify five subgroups of fourth graders based on health behaviors hypothesized to function as mechanisms responsible for the association between EF and obesity, including high-fat/high-sugar food intake, sedentary behavior, weight consciousness, dieting, physical activity, and healthy eating behaviors. A cross-sectional follow-up study with this same sample demonstrated that EF deficits were associated with membership in the most unhealthy latent classes (11).
In the current paper, we took a longitudinal, person-centered approach, multidimensional growth mixture modeling (MGMM) (12,13), to identify unique classes of conjoint development of EF and BMI from ages 3 to 12 years among a sample at high risk for obesity. MGMM is an innovative approach that identifies heterogeneous classes or patterns of development in multiple parallel processes. This approach is advantageous over variable-centered approaches (e.g., cross-lagged analyses) for identifying and visualizing patterns of change in processes that are meaningful, data driven, and easy to interpret. We hypothesized the following: (1) that distinctive classes of EF and BMI development would be identified; (2) that classes characterized by EF proficiency would also display obesity resilience (a nonobesity BMI trajectory); (3) that classes characterized by deficits in EF development would display patterns of rapid BMI increases; and (4) that obesity-related risk factors would predict membership in classes characterized by EF deficits and rapid BMI increases. Indicators of obesity predisposition included parental weight status and infant growth patterns, and we also measured risk due to early exposure to adversity (e.g., poverty).
Methods
Participants
Data were drawn from the Family Life Project (FLP) (14), a longitudinal study designed to study a representative sample of families residing in predominantly low-income, nonurban communities in the United States. Mothers residing in eastern North Carolina and central Pennsylvania were recruited at the time of childbirth, and low-income and African American families were oversampled. Youth (n = 1,292) were followed from birth to age 12 years, with home visits occurring at ages 2 and 6 months and ages 2, 3, 4, 5, 7, and 12 years. Extensive study details are provided at https://flp.fpg.unc.edu. Because we were interested in modeling normative growth patterns of BMI in childhood between ages 3 and 12 years, we excluded youth with < 2 BMI data points (n = 157) and those whose gestational age was less than 37 weeks or greater than 42 weeks (n = 92). Youth with missing EF data between ages 3 and 5 years were not included in the analyses (n = 171). This resulted in a final sample of 948 youth. Included youth did not differ from those excluded (n = 344) in ethnicity (White vs. racial/ethnic minority; P = 0.916), sex (P = 0.445), state of residence (P = 0.191), maternal education (P = 0.755), or income to needs ratio (INR; P = 0.057).
The final sample was 51.5% male, 55.3% White, and 41.7% African American. At the 2-month visit, average household income was $35,789 (SD = $30,553; median = $28,736) and INR was 1.85 (SD = 1.73; median = 1.42). Over one in three households (35.9%) were below the poverty line with INR < 1.0, while 15.6% were just above the poverty line with INR between 1.0 and 1.5, and 13.8% were near poor with INR between 1.5 and 2.0. At study recruitment, 14.9% of mothers reported having completed a 4-year college education, and 50.1% reported having completed high school/general equivalency diploma. Given the presence of the obesity-related risk factors (e.g., high poverty rates, low education levels) (15), residence in rural communities (16), and the high prevalence of parental overweight and obesity (65.8% for mothers and 69.4% for fathers at the 2-month assessment) (17), the youth in this sample were considered to be at high risk for overweight and obesity. This study was approved by the institutional review board at the sponsoring universities.
Measures
Child anthropometrics.
Weight and height were measured once by trained interviewers (heavy clothing and shoes were removed prior to measurements). At ages 2 and 6 months, youth weight (0.01 kg) and recumbent length (0.1 cm) were used to compute weight-for-length and BMI percentiles based on World Health Organization guidelines (18). Between ages 2 and 12 years, weight (0.1 kg) and standing height (0.1 cm) were used to compute age- and sex-specific BMI scores and percentiles based on Centers for Disease Control and Prevention guidelines (19). Overweight was defined as BMI ≥ 85th percentile, and three categories of obesity were classified as follows: class I obesity as BMI ≥ 95th percentile; class II obesity as BMI ≥ 120% of the 95th percentile (severe obesity); and class III obesity as BMI ≥ 140% of the 95th percentile (severe obesity).
EF.
A battery of EF tasks was administered at ages 3, 4, and 5 years. The battery included three inhibitory tasks (a Simon-like spatial conflict task, a Stroop-like silly sounds task, and the farm animal go-no-go task), two working memory tasks (a span-like task and a self-ordered pointing task), and one cognitive flexibility task modeled on the Dimensional Change Card Sort task (20). At age 3 years, a similar (but not identical) spatial conflict task was used, and the self-ordered pointing task was not administered; remaining tasks were unchanged over time. Instrument scores were converted to a common longitudinal scale, in which scores were scaled to mean = 0 and SD = 1, and performance at age 4 years was the reference. Given that EF increases with age (4), instrument scores at age 3 years were lower than at age 4 years and likely negative values, whereas scores at age 5 years were higher than at age 4 years and likely positive values. Item response theory was used to calculate expected EF scores, as this has been shown to more accurately and precisely measure an individual’s latent performance ability compared with classical test theory (20). Raw mean EF scores, representing the percent of correct responses in the tasks (range: 0–1), were also calculated to permit comparisons between EF scores in the current sample with previous work (21). Task procedures, psychometrics, and scoring are presented elsewhere (20,22). As is typical of EF measures, reliability coefficients for the EF composites were relatively low, α range = 0.37 to 0.55.
Obesity predisposition.
Parental BMI and infant birth weight and early growth were used as crude, yet reliable, indicators of obesity predisposition because of genetics and shared environment (23), as well as a propensity toward rapid weight gain in early childhood (24). At age 2 months, mothers recalled their child’s birth weight in pounds and ounces (converted to grams). At the child’s 2-year-old check-in, mothers’ heights and weights were measured once by trained research assistants. Mothers self-reported their height and weight at the 2- and 6-month-old check-ins and reported the father’s height and weight at the 2-month-old check-in. Adult overweight was defined as BMI of 25.0 to < 30.0, and three categories of adult obesity were defined as follows: class I obesity as BMI of 30.0 to < 35.0; class II obesity as BMI of 35.0 to < 40.0; and class III obesity as BMI ≥ 40 (severe obesity).
Cumulative risk.
We employed a cumulative risk score previously created and validated within the FLP sample and described elsewhere (14). The score comprises the following seven indicators: maternal education, employment hours, job prestige, household density, household INR, constant spouse/partner living in the home, and neighborhood noise and safety. Each indicator was assessed at 6 and 15 months and again at 2 and 3 years. Household income was based on anyone who resided in the home ≥ 3 nights/week. Based on household income and size, INR was calculated using yearly poverty threshold values. At the end of each visit, trained research assistants rated each family’s neighborhood in terms of noise, safety, and community safety, and the ratings (1 to 4) were then averaged. Risk indicators having a positive orientation (e.g., maternal education) were reverse coded, and each risk indicator was then standardized and averaged across time points to create aggregate risk indicators. Correlations between aggregate risk indicators ranged from 0.12 to 0.59 for the full FLP sample (14). Aggregate risk indicators were averaged to create a cumulative risk score, representing exposure to adversity between ages 6 months to 3 years.
Covariates.
Youth IQ at age 3 years was assessed using receptive verbal ability and block design subscales of the Wechsler Preschool and Primary Scales of Intelligence (25). Child sex, race, and state of residence were included as covariates.
Statistical analysis
Analyses were performed in Mplus 8.0 (Muthen & Muthen, Los Angeles, California), which allows for the estimation of models with missing data (at random) using full information maximum likelihood. As suggested in Jung and Wickrama (26), first, unconditional latent growth curve models were fitted to determine whether change in EF from ages 3 to 5 years was linear and whether change in BMI from ages 3 to 12 years was linear, quadratic, and/or cubic. Second, conditional latent growth curve models were conducted in which latent intercept and slope growth terms were regressed on covariates (i.e., child sex [1 = male, 0 = female]; race [1 = Black, 0 = non-Black]; IQ; and state residence [1 = Pennsylvania, 0 = North Carolina]) individually. Third, MGMM was used to identify trajectories (or classes) of change in BMI and EF. This approach has been used elsewhere to identify patterns of change in multiple, overlapping processes (e.g., chronic illness and health behaviors) (13). To simplify model computation and reduce convergence issues, within-class linear and quadratic variances were constrained to zero; intercept variances were freely estimated. Latent intercepts and class membership were regressed on covariates. One to six classes were estimated. Model fit was assessed using the Bayesian Information Criterion (BIC) and Lo-Mendell-Rubin Likelihood Ratio Test (LMR-LRT). Lower BIC values indicate models with better fit, while LMR-LRT evaluates whether a model with k classes is better than a model with k − 1 classes. We also evaluated entropy and model interpretability. Lastly, weighted mean difference testing was performed using the AUXILLIARY BCH option to evaluate whether parental BMI, infant birth weight and growth, and cumulative risk predicted class membership. BMI weight classifications (e.g., overweight) by class membership were reported using available data per time point.
Results
Distribution of BMI growth and EF data for each time of measurement is shown in Table 1. On average, raw EF scores in the current sample increased from 49% to 73% from ages 3 to 5 years, which is similar to scores previously reported elsewhere (21).
TABLE 1.
Distribution of BMI growth and EF by time of measurement between 3 and 12 years of age
| Mean | SD | N | |
|---|---|---|---|
| BMI | |||
| 3 years | 16.6 | 1.6 | 915 |
| 4 years | 16.5 | 1.9 | 853 |
| 5 years | 16.6 | 2.3 | 870 |
| 7 years | 17.4 | 3.4 | 855 |
| 12 years | 23.5 | 6.4 | 657 |
| EF | |||
| 3 years | −0.52 | 0.54 | 874 |
| 4 years | −0.11 | 0.51 | 864 |
| 5 years | 0.32 | 0.47 | 876 |
EF scores were assessed using the EF Touch battery of tasks and computed using Item Response Theory. Instrument scores were converted to a common longitudinal scale, in which scores were scaled to SD = 0 and SD = 1, and age 4 performance was the reference.
EF, executive function.
EF-BMI trajectories
For the BMI data, both the linear (B = −1.475, P < 0.001) and quadratic slopes (B = 5.853, P < 0.001) differed from zero, indicating that youth displayed accelerated physical growth from ages 3 to 12 years. For the EF data, the linear slope was significantly different from zero (B = 3.75, P < 0.001), indicating that average EF increased from 3 to 5 years. Higher EF at 3 years (i.e., intercept) was associated with being female (β = −0.237, P < 0.001) and non-Black (β = 0.487, P < 0.001), having a higher IQ ( = 0.767, P < 0.001), and residing in Pennsylvania (β = 0.452, P < 0.001). Child BMI at age 3 years was unrelated to study covariates.
MGMM fit statistics are shown in Table 2. The three- and four-class models had very similar, low BIC values. The LMR-LRT indicated that the three-class model fit the data better than a model with one less class, but this index was not statistically significant for the four-class model. Both the three- and four-class models identified the same three classes; however, the four-class model also identified a class of 45 youth (mostly without overweight) who displayed an outlier trajectory of very low EF scores with little to no change from ages 3 to 5 years (−0.86 to −0.70, respectively), in comparison with the other classes (Figure 1). Taken together, a three-class model was selected.
TABLE 2.
Model fit statistics from MGMM
| Latent classes | BIC | Convergence | LMR-LRT | Entropy | Size of smallest class (n) |
|---|---|---|---|---|---|
| 1 | 19,199 | Yes | NA | 100.0 | 948 (100%) |
| 2 | 18,589 | Yes | 0.0016 | 0.913 | 116 (12.2%) |
| 3 | 18,468 | Yes | 0.0390 | 0.889 | 52 (5.5%) |
| 4 | 18,460 | Yes | 0.1097 | 0.858 | 47 (5.0%) |
| 5 | 18,503 | Yes | 0.4949 | 0.667 | 44 (4.6%) |
BIC, Bayesian Information Criterion; LMR-LRT, Lo-Mendell-Rubin Likelihood Ratio Test; MGMM, multidimensional growth mixture modeling.
Figure 1.
Multidimensional growth mixture modeling solution of executive function (EF)-BMI trajectory latent classes plotted using mean (SE) executive function scores. Different alphanumeric subscripts within the same time point indicate mean differences between groups at P < 0.05, found in weighted mean difference testing performed in Mplus 8.0 using the AUXILLIARY BCH option.
To assist with class interpretation, the change in EF, BMI percentiles, and percentage of the 95th BMI percentile by class membership were plotted in Figure 1 and Figure 2A and 2B, respectively. The majority of youth belonged to the “High EF – High Obesity Resilience” class (n = 807, 85.1%). On average, this class had the highest levels of EF from ages 3 to 5 years and BMI percentiles that ranged between 58.9 and 64.7 from ages 3 to 12 years. As shown in Table 3, 77.5% and 62.8% of youth in this class were without overweight at ages 3 and 12 years, respectively. The remaining two classes displayed similar, low levels of EF (Figure 1) and high proportions of severe obesity (or youth class II obesity) at age 12 years (Table 3); however, the two classes differed in their path to severe obesity. The “Low EF – Early-Onset Severe Obesity” class (n = 52, 5.5%) exhibited a rapid path to obesity and exceeded the cutoff for severe obesity by age 5 years, while the “Low EF – Delayed-Onset Severe Obesity” class (n = 89, 9.4%) showed evidence of severe obesity at age 12. Of note, 70.0% of children who developed moderate obesity at age 12 years were classified into the high EF, high resilience class and 30.0% into the low EF, severe obesity classes, indicating that children with moderate obesity (class I) had greater variance in EF levels.
Figure 2.
Multidimensional growth mixture modeling solution of EF-BMI trajectory latent classes plotted using mean (SE) BMI percentiles and percentages of the 95th percentile. BMI weight classification cutoffs (i.e., solid blocks) were determined based on Centers for Disease Control and Prevention recommendations for the three categories of obesity: class I obesity as ≥ 95th BMI percentile, class II obesity (severe obesity) as ≥ 120% of the 95th BMI percentile, and class III (severe obesity) as ≥ 140% of the 95th percentile. Different alphanumeric subscripts (i.e., a, b, and c) within the same time point indicate mean differences between groups at P < 0.05, found in weighted mean difference testing performed in Mplus 8.0 using the AUXILLIARY BCH option.
TABLE 3.
Percentages of weight classifications by EF-BMI trajectory class membership
| High EF – High Obesity Resilience (n = 807, 85.1%) | Low EF – Delayed-Onset Severe Obesity (n = 89, 9.4%) | Low EF – Early-Onset Severe Obesity (n = 52, 5.5%) | Total sample | |
|---|---|---|---|---|
| (N = 948, 100.0%) | ||||
| Age 3 | ||||
| Underweight, % | 2.3 | 1.2 | 0.0 | 2.1 |
| No overweight, % | 77.5 | 39.1 | 36.0 | 71.6 |
| Overweight, % | 15.5 | 18.4 | 16.0 | 15.8 |
| Obesity, % | 4.6 | 34.5 | 34.0 | 9.0 |
| Severe obesity, % | 0.1 | 6.8 | 14.0 | 1.5 |
| Age 4 | ||||
| Underweight, % | 2.6 | 0.0 | 2.0 | 2.3 |
| No overweight, % | 75.7 | 23.8 | 8.0 | 66.9 |
| Overweight, % | 16.0 | 23.8 | 6.0 | 16.1 |
| Obesity, % | 5.7 | 47.5 | 40.0 | 11.6 |
| Severe obesity, % | 0.0 | 4.9 | 44.4 | 3.1 |
| Age 5 | ||||
| Underweight, % | 2.4 | 0.0 | 0.0 | 2.1 |
| No overweight, % | 76.3 | 16.5 | 2.0 | 66.1 |
| Overweight, % | 14.5 | 27.1 | 3.9 | 15.1 |
| Obesity, % | 6.7 | 45.9 | 33.3 | 12.0 |
| Severe obesity, % | 0.1 | 10.5 | 60.8 | 3.7 |
| Age 7 | ||||
| Underweight, % | 2.9 | 0.0 | 0.0 | 2.5 |
| No overweight, % | 77.8 | 7.3 | 0.0 | 66.6 |
| Overweight, % | 13.8 | 20.7 | 2.0 | 13.8 |
| Obesity, % | 5.4 | 50.0 | 18.4 | 10.4 |
| Severe obesity, % | 0.1 | 22.0 | 79.6 | 6.7 |
| Age 12 | ||||
| Underweight, % | 3.2 | 0.0 | 0.0 | 2.6 |
| No overweight, % | 62.8 | 0.0 | 2.2 | 51.4 |
| Overweight, % | 20.6 | 0.0 | 4.4 | 17.1 |
| Obesity, % | 13.2 | 20.7 | 31.1 | 15.3 |
| Severe obesity, % | 0.2 | 79.3 | 62.3 | 13.6 |
BMI weight classifications were determined based on Centers for Disease Control and Prevention recommendations for underweight (BMI < 5th percentile), overweight (BMI ≥ 85th percentile), and three categories of obesity: class I obesity as ≥ 95th BMI percentile, class II obesity (severe obesity) as ≥ 120% of the 95th BMI percentile, and class III (severe obesity) as ≥ 140% of the 95th percentile.
EF, executive function.
Genetic predisposition
As shown in Table 4, the low EF, early-onset class had the highest birth weight and higher BMI and weight-for-length at age 6 months, compared with the other classes. However, the low EF, delayed-onset class had greater increases in BMI between ages 2 and 24 months, relative to the high EF, high resilience class, such that by age 2 years, both of the low EF, severe obesity classes displayed similar high risk for obesity at the 83rd BMI percentile. Compared with the high EF, high resilience class, both low EF, severe obesity classes had parents with high BMI (near or in the obesity range ≥ 30) at all time points between ages 2 and 24 months. Low EF, delayed-onset was the only class for which mothers’ average BMI at the 24-month visit would classify them as class II obesity. Although the high EF, high resilience class had parents with lower BMI than the other classes, these parents had BMI that still classified them as having overweight or obesity at all time points between ages 2 and 24 months.
TABLE 4.
Mean (SE) differences1 in study variables by EF-BMI trajectory class membership
| High EF – High Obesity Resilience (n = 807, 85.1%) | Low EF – Delayed Onset Severe Obesity (n = 89, 9.4%) | Low EF – Early Onset Severe Obesity (n = 52, 5.5%) | F | |
|---|---|---|---|---|
| Early life growth | ||||
| Birth weight (kg) | 3.4 ± 0.0a | 3.3 ± 0.1a | 3.5 ± 0.1b2 | 4.79 |
| BMI, 2 months | 16.3 ± 0.8a | 16.6 ± 0.5ab | 17.3 ± 0.2b | 14.35*** |
| BMI, 6 months | 17.8 ± 0.1a | 18.1 ± 0.2a | 18.9 ± 0.3b | 16.03*** |
| BMI, 24 months | 17.1 ± 0.1a | 18.9 ± 0.3b | 18.5 ± 0.3b | 75.17*** |
| BMI change, 2 to 24 months | 0.8 ± 0.1a | 2.3 ± 0.5b | 1.3 ± 0.4ab | 14.35*** |
| WFL, 2 months | 53.7 ± 1.2a | 52.7 ± 3.8ab | 68.2 ± 4.2b | 11.18** |
| WFL, 6 months | 62.9 ± 1.0a | 72.0 ± 3.4b | 82.4 ± 3.5c | 34.30*** |
| BMI percentile, 24 months | 61.6 ± 1.0a | 83.4 ± 2.9b | 83.3 ± 2.7b | 99.61*** |
| Parental BMI | ||||
| Mom BMI, 2 months | 27.3 ± 0.2a | 33.2 ± 1.1b | 33.3 ± 1.1b | 54.88*** |
| Mom BMI, 6 months | 30.4 ± 0.7a | 35.9 ± 2.3b | 34.9 ± 1.1b | 14.8*** |
| Mom BMI, 24 months | 30.6 ± 1.0a | 42.4 ± 4.9b | 37.5 ± 2.8b | 10.9** |
| Dad BMI, 2 months | 26.7 ± 0.2a | 29.3 ± 0.7b | 30.5 ± 1.1b | 10.9** |
| Cumulative risk | ||||
| Cumulative risk, averaged 6 months to 3 years | −0.04 ± 0.03a | 0.18 ± 0.08b3 | −0.06 ± 0.11a | 7.25* |
| Child characteristics | ||||
| Child sex (male = 1), % | 52.8 ± 1.8a | 30.9 ± 5.9b | 49.3 ± 7.6ab | 12.1** |
| Child race (Black = 1), % | 39.3 ± 1.8a | 69.3 ± 5.9b | 36.1 ± 7.4a | 22.25*** |
Different alphanumeric subscripts within the same row indicate mean differences between groups at P < 0.05, found in weighted mean difference testing performed in Mplus 8.0 using the AUXILLIARY BCH option.
Mean difference between the “Low EF – Delayed-Onset Severe Obesity” class and “Low EF – Early-Onset Severe Obesity” class was marginally significant at P = 0.057. The “High EF – High Obesity Resilience” class and “Low EF – Early-Onset Severe Obesity” class were statistically significant different at P < 0.05.
Mean difference between the “Low EF – Delayed-Onset Severe Obesity” class and “Low EF – Early-Onset Severe Obesity” class was marginally significant at P = 0.080. The “High EF – High Obesity Resilience” class and “Low EF – Delayed-Onset Severe Obesity” class were statistically significantly different at P < 0.05.
P > 0.05.
P > 0.01.
P > 0.001.
EF, executive function; WFL, weight-for-length.
Cumulative risk
As shown in Table 4, the low EF, delayed-onset class displayed higher levels of cumulative risk than the high EF, high resilience (P < 0.05) and low EF, early-onset (P = 0.080) classes. Additionally, the low EF, delayed-onset class had more Black youth than the other two classes and more female youth than the high EF, high resilience class.
Discussion
We examined the longitudinal, conjoint development of EF (ages 3 to 5 years) and BMI trajectories (ages 3 to 12 years) in childhood using MGMM, a person-centered approach. To the best of our knowledge, this is the first study to demonstrate this association using longitudinal data between infancy and adolescence. Consistent with our hypotheses and previous work (27), we identified a group of youth displaying higher EF and obesity resilience. However, we also saw that this group contained a large portion of children with moderate levels of obesity. Although this was not expected, there is some evidence to suggest that high levels of EF may not confer the same degree of health protection for Black children compared with White children (28). More research is needed in this area, as the relation between EF and obesity may be more complex for certain demographics. We also hypothesized that children with EF deficits would show the steepest weight gain trajectories. We uniquely identified two low-EF groups with two distinct severe obesity BMI pathways and patterns of obesity-related risk factors. The group with low EF and early-onset severe obesity were the heaviest at birth and in infancy, and more than 60% of the youth in this class had severe obesity by age 5 years. In contrast, the group with lower levels of EF and delayed-onset severe obesity had similar birth weight as the “High EF – High Obesity Resilience” group but then displayed more rapid growth between ages 2 and 24 months and again between ages 7 and 12 years. By age 12 years, 79.3% of youth in this group had severe obesity. Also, 30% of the children with moderate obesity at age 12 years were classified in the severe obesity groups, suggesting that some children with this weight status may display EF deficits. Our findings provide novel longitudinal evidence linking early EF to early growth patterns and suggest that, among youth with low EF, early and middle childhood are critical periods for obesity and severe obesity intervention. We also found that the low-EF youth had parents with very high BMI, and the delayed-onset group displayed the highest levels of cumulative risk and were most likely to be female and Black.
Longitudinal evidence suggests that early EF, particularly measures of inhibitory control, precedes obesity development in childhood (29,30). Researchers have proposed that EF processes may be indicators of biobehavioral dysregulation in other domains of development that confer risk for obesity, such as dysregulated eating behaviors (31), or through appetitive behaviors related to overeating and greater energy intake (32,33). The relation between EF and obesity may also be bidirectional, as obesity is associated with increased neuroinflammation leading to decreased synaptic spine density and cognitive deficits (34). It has also been proposed that individual variability in the energy demand of brain development in early childhood may lead to individual variability in adiposity gains during this period (29,35). Our findings indicate that EF may precede the development of severe obesity for some youth, but not for others. Despite the delayed-onset group having similar birth weight as the obesity-resilient group, all youth in the delayed-onset group developed obesity or severe obesity by age 12 years. This is in contrast with the early-onset group, whose higher birth weight and elevated weight status were evident before EF processes developmentally come online (ages 3 to 5 years). The unique “early-onset” pattern of rapid weight gain may be a product of early programming, as indicated by heavier birth weight and infancy weight-for-stature. There may also be other, unmeasured risk factors (e.g., genetic or biobehavioral factors) that are stronger drivers of accelerated growth than EF for this group, or it could be that excess adiposity during early childhood facilitated the development of cognitive deficits via inflammation, in addition to possible poor eating and sedentary behaviors.
The delayed-onset group displayed a well-documented profile of obesity risk factors, including infant rapid weight gain (24), having a mother with class I or II obesity at age 24 months (23), early exposure to cumulative risk (36), and being Black (37). Given that EF appears to mediate the association between cumulative risk and obesity risk (38), it is possible that this profile of risk factors contributed to the high proportions of severe obesity within this group. There may also be unmeasured obesity-protective or -resilience factors that contributed to delayed development of severe obesity, such as children’s appetitive behaviors, food preferences, and/or physical activity patterns. Few studies have examined factors that predict healthy weight or weight gain in children, but those that have done so have examined factors such as parent involvement in reducing children’s physical and/or sedentary activity patterns (39), child dietary patterns (40), and family resilience variables related to obesity (41) (such as child-feeding practices across childhood).
Given our high-risk sample of youth exposed to long-term adversity, proficient EF may be conceptualized as a measure of resilience, which is described as the capacity to adjust, adapt, self-stabilize, and self-regulate in the context of adversity (42). Findings from a nationally representative study of youth aged 10 to 17 years found that child resilience, measured using indicators of curiosity, completing tasks, and staying calm/controlled when challenged, was associated with a lower risk for obesity (43), and that the relation was strongest among youth living in high-poverty households, which has tremendous implications for prevention. Our findings provide evidence that EF may confer a degree of obesity resilience for some youth exposed to poverty. Furthermore, given the high rates of parental overweight and obesity in this sample, the majority of youth had a high predisposition for obesity. The identification of a class of predominantly obesity-resilient youth with high levels of EF suggests that efforts to improve EF in early childhood may be an effective approach for obesity prevention in youth at high risk for obesity.
There are several limitations. Given that the FLP was not designed to examine childhood obesity, we are limited in our ability to assess potential obesogenic mechanisms (e.g., dietary patterns) that may explain the relationship between EF and obesity. We used a composite measure of EF, which precludes us from drawing conclusions regarding specific aspects of EF (e.g., inhibitory control, working memory). There may also be unmeasured EF processes (e.g., decision-making) that better predict obesity. We were limited in our ability to describe EF changes through age 12 years because we did not have consistent EF measures after age 5 years. In addition, our EF measures demonstrated low reliability scores, which may have introduced noise into our analyses; however, our reliability scores are consistent with other studies of EF among preschool children. Our findings cannot be generalized to all rural youth in the US. Lastly, mothers reported on youth birth weight, parental height and weights, and cumulative risk factors, which may be subject to potential bias. Furthermore, there is a possibility for residual confounding of factors, such as maternal prepregnancy BMI, smoking during pregnancy, and breastfeeding duration. Despite these limitations, our study contributes new information on longitudinal relations between EF and growth in a predominantly rural, low-income sample at high risk for obesity. Our use of MGMM allowed us to identify unique relations between EF and BMI change over time, and it is likely that the low-EF, early- and late-onset severe obesity groups would not have been identified with conventional approaches often used to describe change in BMI (e.g., change scores).
In conclusion, our findings demonstrate that youth with EF deficits in early childhood are at the greatest risk for the development of severe obesity by age 12 years. Interventions designed to train youth to improve their EF have been shown to be efficacious in reducing obesogenic behaviors (44,45). Notably, the majority of the sample demonstrated higher levels of EF and were largely obesity resilient, highlighting EF as a potential marker of resilience that may contribute to obesity resilience. However, children with moderate obesity displayed both low and high EF levels, which suggests that more work needs to be done to better understand within-group factors that explain the EF–obesity link.
Study Importance.
What is already known?
Both pediatric obesity and executive function develop rapidly in early childhood.
Executive function deficits have been linked to pediatric obesity; however, most of this research is cross-sectional or utilized variable-centered methods, which has limited our understanding of the link.
What does this study add?
The majority of youth exhibited obesity resilience, characterized by high executive function and nonobesity between 3 to 12 years.
The remaining two groups displayed EF deficits with two unique trajectories of rapid BMI change and patterns of obesity risk factors, suggesting that the obesity-executive function link is dynamic.
Acknowledgments
We would like to express our gratitude to the FLP families and home visit data collectors who contributed to this study.
Funding agencies:
This study utilizes data from the Family Life Project (FLP) (https://flp.fpg.unc.edu/). Grant support for FLP data collection was provided by the National Institute of Child Health and Human Development grants R01 HD081252 and P01 HD039667. Grant support for manuscript preparation was provided by National Institute of Child Health and Human Development grants R01 HD074807.
Footnotes
Disclosure: All authors declared no conflict of interest.
References
- 1.Ogden CL, Fryar CD, Hales CM, et al. Differences in obesity prevalence by demographics and urbanization in US children and adolescents, 2013–2016. JAMA 2018;319:2410–2418. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Riggs NR, Greenberg MT, Rhoades B. Early risk for problem behavior and substance use: targeted interventions for the promotion of inhibitory control. In: Bardo MT, Fishbein DH, Milich R, eds. Inhibitory Control and Drug Abuse Prevention: From Research to Translation. Springer, New York; 2011:249–262. [Google Scholar]
- 3.Werchan DM, Amso D. A novel ecological account of prefrontal cortex functional development. Psychol Rev 2017;124:720–739. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Zelazo PD, Carlson SM, Kesek A. Development of executive function in childhood. In: Nelson CA, Luciana M, eds. Handbook of Developmental Cognitive Neuroscience. MIT Press; 2008:553–574. [Google Scholar]
- 5.Davidson MC, Amso D, Anderson LC, Diamond A. Development of cognitive control and executive functions from 4 to 13 years: evidence from manipulations of memory, inhibition, and task switching. Neuropsychologia 2006;44:2037–2078. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Riggs N, Chou CP, Spruijt-Metz D, Pentz MA. Executive cognitive function as a correlate and predictor of child food intake and physical activity. Child Neuropsychology 2010;16:279–292. [DOI] [PubMed] [Google Scholar]
- 7.Reinert KR, Po’e EK, Barkin SL. The relationship between executive function and obesity in children and adolescents: a systematic literature review. J Obes 2013;2013:820956. doi: 10.1155/2013/820956 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Riggs NR, Greenberg MT, Kusché CA, Pentz MA. The mediational role of neurocognition in the behavioral outcomes of a social-emotional prevention program in elementary school students: effects of the PATHS curriculum. Prev Sci 2006;7:91–102. [DOI] [PubMed] [Google Scholar]
- 9.Berge JM, Wall M, Bauer KW, Neumark-Sztainer D. Parenting characteristics in the home environment and adolescent overweight: a latent class analysis. Obesity (Silver Spring) 2010;18:818–825. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Huh J, Riggs NR, Spruijt-Metz D, et al. Identifying patterns of eating and physical activity in children: a latent class analysis of obesity risk. Obesity (Silver Spring) 2011;19:652–658. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Riggs NR, Huh J, Chou C-P, Spruijt-Metz D, Pentz MA. Executive function and latent classes of childhood obesity risk. J Behav Med 2012;35:642–650. [DOI] [PubMed] [Google Scholar]
- 12.Wickrama KAS, Lee TK, O’Neal CW, Lorenz FO. Higher-Order Growth Curves and Mixture Modeling With Mplus: A Practical Guide. Routledge/Taylor & Francis Group; 2016. [Google Scholar]
- 13.Wickrama K, Mancini JA, Kwag K, Kwon J. Heterogeneity in multidimensional health trajectories of late old years and socioeconomic stratification: a latent trajectory class analysis. J Gerontol B Psychol Sci Soc Sci 2012;68:290–297. [DOI] [PubMed] [Google Scholar]
- 14.Vernon-Feagans L, Cox M; FLF Key Investigators. The Family Life Project: an epidemiological and developmental study of young children living in poor rural communities. Monogr Soc Res Child Dev 2013;78:1–150. [DOI] [PubMed] [Google Scholar]
- 15.Ogden CL, Lamb MM, Carroll MD, Flegal KM. Obesity and socioeconomic status in children and adolescents: United States, 2005–2008. NCHS Data Brief, no, 51. National Center for Health Statistics; 2010. [PubMed] [Google Scholar]
- 16.Johnson JA 3rd, Johnson AM. Urban-rural differences in childhood and adolescent obesity in the United States: a systematic review and meta-analysis. Child Obes 2015;11:233–241. [DOI] [PubMed] [Google Scholar]
- 17.Zalbahar N, Najman J, McIntyre HD, Mamun A. Parental pre-pregnancy obesity and the risk of offspring weight and body mass index change from childhood to adulthood. Clin Obes 2017;7:206–215. [DOI] [PubMed] [Google Scholar]
- 18.World Health Organization. WHO Child Growth Standards: Head Circumference-for-Age, Arm Circumference-for-Age, Triceps Skinfold-for-Age and Subscapular Skinfold-for-Age: Methods and Development. WHO; 2007. [Google Scholar]
- 19.Kuczmarski RJ, Ogden CL, Guo SS, et al. 2000 CDC Growth Charts for the United States: Methods and Development. Vital and Health Statistics, series 11, no, 246. National Center for Health Statistics; 2002. [PubMed] [Google Scholar]
- 20.Willoughby MT, Wirth R, Blair CB. Contributions of modern measurement theory to measuring executive function in early childhood: an empirical demonstration. J Exp Child Psychol 2011;108:414–435. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Camerota M, Willoughby MT, Kuhn LJ, Blair CB. The Childhood Executive Functioning Inventory (CHEXI): factor structure, measurement invariance, and correlates in US pre-schoolers. Child Neuropsychol 2018;24:322–337. [DOI] [PubMed] [Google Scholar]
- 22.Willoughby MT, Wirth R, Blair CB. Executive function in early childhood: longitudinal measurement invariance and developmental change. Psychol Assess 2012;24:418–431. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Whitaker KL, Jarvis MJ, Beeken RJ, Boniface D, Wardle J. Comparing maternal and paternal intergenerational transmission of obesity risk in a large population-based sample. Am J Clin Nutr 2010;91:1560–1567. [DOI] [PubMed] [Google Scholar]
- 24.Dennison BA, Edmunds LS, Stratton HH, Pruzek RM. Rapid infant weight gain predicts childhood overweight. Obesity (Silver Spring) 2006;14:491–499. [DOI] [PubMed] [Google Scholar]
- 25.Wechsler D Wechsler Primary and Preschool Scale of Intelligence. The Psychological Corporation; 2002. [Google Scholar]
- 26.Jung T, Wickrama K. An introduction to latent class growth analysis and growth mixture modeling. Soc Pers Psychol Compass 2008;2:302–317. [Google Scholar]
- 27.Liang J, Matheson BE, Kaye WH, Boutelle KN. Neurocognitive correlates of obesity and obesity-related behaviors in children and adolescents. Int J Obes (Lond) 2014;38:494–506. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Miller GE, Yu T, Chen E, Brody GH. Self-control forecasts better psychosocial outcomes but faster epigenetic aging in low-SES youth. Proc Natl Acad Sci 2015;112:10325–10330. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Blair C, Kuzawa CW, Willoughby MT. The development of executive function in early childhood is inversely related to change in body mass index: evidence for an energetic tradeoff? Dev Sci 2020;23:e12860. doi: 10.1111/desc.12860 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Datar A, Chung PJ. Childhood self-control and adolescent obesity: evidence from longitudinal data on a national cohort. Child Obes 2018;14:238–247. [DOI] [PubMed] [Google Scholar]
- 31.Francis LA, Riggs NR. Executive function and self-regulatory influences on children’s eating. In: Lumeng JC, Fisher JO, eds. Pediatric Food Preferences and Eating Behaviors. Academic Press; 2018:183–206. [Google Scholar]
- 32.Tan CC, Lumeng JC. Associations between cool and hot executive functions and children’s eating behavior. Curr Nutr Rep 2018;7:21–28. [DOI] [PubMed] [Google Scholar]
- 33.Goldschmidt AB, O’Brien S, Lavender JM, et al. Executive functioning in a racially diverse sample of children who are overweight and at risk for eating disorders. Appetite 2018;124:43–49. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Bocarsly ME, Fasolino M, Kane GA, et al. Obesity diminishes synaptic markers, alters microglial morphology, and impairs cognitive function. Proc Natl Acad Sci 2015;112:15731–15736. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Kuzawa CW, Blair C. A hypothesis linking the energy demand of the brain to obesity risk. Proc Natl Acad Sci U S A 2019;116:13266–13275. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Miller AL, Lumeng JC. Pathways of association from stress to obesity in early childhood. Obesity (Silver Spring) 2018;26:1117–1124. [DOI] [PubMed] [Google Scholar]
- 37.Skinner AC, Ravanbakht SN, Skelton JA, Perrin EM, Armstrong SC. Prevalence of obesity and severe obesity in US children, 1999–2016. Pediatrics 2018;141:e20173459. doi: 10.1542/peds.2017-3459 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Evans GW, Fuller-Rowell TE, Doan SN. Childhood cumulative risk and obesity: the mediating role of self-regulatory ability. Pediatrics 2012;129:e68–e73. [DOI] [PubMed] [Google Scholar]
- 39.Moore LL, Gao D, Bradlee ML, et al. Does early physical activity predict body fat change throughout childhood? Prev Med 2003;37:10–17. [DOI] [PubMed] [Google Scholar]
- 40.Haines J, Neumark-Sztainer D, Wall M, Story M. Personal, behavioral, and environmental risk and protective factors for adolescent overweight. Obesity (Silver Spring) 2007;15:2748–2760. [DOI] [PubMed] [Google Scholar]
- 41.Yang Y, Shields GS, Guo C, Liu Y. Executive function performance in obesity and overweight individuals: a meta-analysis and review. Neurosci Biobehav Rev 2018;84:225–244. [DOI] [PubMed] [Google Scholar]
- 42.MacPhee D, Lunkenheimer E, Riggs N. Resilience as regulation of developmental and family processes. Fam Relat 2015;64:153–175. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Foster BA, Weinstein K. Moderating effects of components of resilience on obesity across income strata in the national survey of children’s health. Acad Pediatr 2019;19:58–66. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Riggs NR, Sakuma K-LK, Pentz MA. Preventing risk for obesity by promoting self-regulation and decision-making skills: pilot results from the PATHWAYS to health program (PATHWAYS). Eval Rev 2007;31:287–310. [DOI] [PubMed] [Google Scholar]
- 45.Verbeken S, Braet C, Goossens L, Van der Oord S. Executive function training with game elements for obese children: a novel treatment to enhance self-regulatory abilities for weight-control. Behav Res Ther 2013;51:290–299. [DOI] [PubMed] [Google Scholar]