Family Psychosocial Assets, Child Behavioral Regulation, and Obesity

Data were drawn from the National Institute of Child Health and Human Development Study of Early Child Care and Youth Development (NICHD SECCYD)^21,22,23,24 ; study details have been previously published.²⁵  A total of 1364 families with full-term, healthy newborns were recruited from hospitals at 10 data collection sites in 1991. Families were excluded if (1) mothers were <18 years of age, (2) newborns had obvious disabilities at birth or stayed in the hospital >7 days after birth, (3) mothers did not sufficiently converse in English, or (4) families anticipated moving from the catchment area within 3 years. Home and laboratory visits were completed at 1, 6, 15, 24, 36, and 54 months and 7, 9, 11, 12, 13, 14, and 15 years of age. Attrition across the course of the study resulted in a sample of 1009 participants at 15 years. Only children having BMI trajectories (described below) were retained in the current study (n = 1077; a loss of 287 participants). As described previously²⁶  and for purposes of modeling normative growth patterns, the 287 participants were excluded for the following reasons: gestational age <37 or >42 weeks (n = 119), <2 BMI data points between 2 and 15 years (n = 165), or implausible BMI data (n = 3). Included youth did not differ from those excluded (n = 287) in race (White versus child of color; P = .142). Included youth were more likely to be girls (P = .052), reside in 2-parent households (P = .007), have higher income-to-needs ratios (P = .010), and have mothers with ≥16 years of education (P = .0001).

We used 4 BMI trajectory groups previously identified in the NICHD SECCYD sample, using latent growth curve modeling as our categorical outcome variable.²⁶  For brevity, children’s weight and standing height were measured in duplicate and used to calculate age- and sex-specific BMI indices.²⁷  Overweight, according to the Centers for Disease Control and Prevention,²⁷  is defined as BMI ≥85th percentile, class I obesity as BMI ≥95th percentile, class II obesity (severe obesity) as BMI ≥120% of 95th percentile, and class III obesity (severe obesity) as BMI ≥140% of 95th percentile. Each BMI trajectory group was defined on the basis of children’s growth from 2 to 15 years. The largest BMI trajectory, referred to as nonoverweight (40th percentile) (n = 468 of 1077; 43.4%), had an average BMI percentile of 40.3 between 2 and 15 years. The second trajectory, nonoverweight (70th percentile) (n = 383; 35.6%), had an average BMI percentile of 72.8 between 2 and 15 years. All youth in these 2 nonoverweight groups had BMI percentiles <85th at each time point. The third trajectory, overweight/obese (n = 157; 14.6%), had average BMI percentiles exceeding the cutoff for overweight between 5 and 15 years, and the cutoff for obesity class I between 11 and 15 years. The final group, severely obese (n = 69, 6.4%), had average BMI percentiles exceeding the cutoff for obesity class II (severe obesity) at all time points between 5 and 15 years, indicating a persistent prevalence of severe obesity.

Table 1 provides the measures, time of assessment, and scoring for each cumulative index of familial assets and risks and child behavioral regulation; the Supplemental Information provides additional details on scoring and psychometrics. For example, the familial assets index (1–15 months) consisted of 3 protective factors (eg, maternal sensitivity) and was created in 3 steps. In step 1, maternal sensitivity scores at 6 and 15 months were averaged for each participant to create a mean score for infancy. In step 2, this mean score was dichotomized; scores in the highest quartile (ie, >75th percentile) were coded as 1; lower scores were coded as 0. Because this measure does not have preexisting cutoffs for high scores, we used quartile splits to identify the highest scores, based on the sample distribution. After these 2 steps were repeated for the remaining 2 familial assets (ie, maternal education, home enrichment), in the third and final step, all 3 dichotomized familial assets were summed to create a cumulative index for infancy (range, 0–3). The same 3-step process was repeated for the familial asset index (24–54 months), familial risk indices (range, 0–6), and child behavioral regulation index (range, 0–3).

At the 1-month home visit, mothers reported household and child demographics (eg, child race/ethnicity). Youth were 77.5% non-Hispanic White; 11.9% were non-Hispanic Black; 5.9% were Hispanic; 3.3% were other; and <3% were American Indian, Asian, or Pacific Islander. Child race/ethnicity was coded as 1 for child of color (22.5%) and 0 for non-Hispanic White (77.5%), given the small proportions of participants of color. Tanner staging criteria³³  were assessed at age 11 years by nurse practitioners to account for the obesity-puberty confounds.³⁴  Boys’ pubertal status was calculated by averaging pubic hair and genital Tanner stages; girls’ pubertal status was calculated by averaging breast and pubic hair Tanner stages. Although puberty data were available at the 9-, 10-, 11-, and 12-year assessments, ages 10 and 11 had the highest level of variability (41.2 and 39.8, respectively); age 11 data were selected because of fewer missing data points (27% vs 33.2%).

Guided by best practices,³⁵  25 multiple imputations were performed by using SAS version 9.4 statistical software (SAS Institute, Cary, NC). All study variables, sociodemographic data, BMI data, and BMI trajectory groups were included (before subscale creation) to inform the imputation. Only 4.9% of data points were missing across the entire sample for the variables of interest. Supplemental Table 12 provides the imputed and nonimputed descriptive statistics, and percent missing for all study variables. The highest level of missing data was observed for pubertal status at 11 years (27.9%); thus, we included pubertal status at 9, 10, 11, and 12 years to inform the imputation. Parameter estimates were combined across imputations using PROC MI ANALYZE. Pearson intercorrelations were run between familial and child cumulative indices. First, bivariate multinomial logistic regression models were run to test for the independent effects of each predictor (ie, familial risk [1–15 months, 24–54 months], familial assets [1–15 months, 24–54 months] child behavioral regulation [36–54 months]) on BMI trajectory membership. Second, multivariate multinomial logistic regression models were run to test whether the indices made unique contributions to predicting BMI trajectory membership as follows: model 1, both familial risk indices; model 2, both familial asset indices; model 3, familial risk and asset indices (1–15 months); and model 4, familial risk and asset indices (24–54 months) and child behavioral regulation. Third, in our final step, we investigated whether familial assets and child behavioral regulation moderated the effects of familial risk on child BMI trajectory membership (models 5–8). Given the small size of the heavier BMI trajectories/groups, moderators showing at least marginal significance (P < .10) were explored. To interpret interaction terms, the sample was divided into groups using median splits on the moderator (eg, low versus high behavioral regulation); multinomial models were run again with this grouping variable as a dichotomous predictor in place of the original continuous variable. All models were run with covariates (eg, child race [1 = child of color, 0 = White], sex [1 = female, 0 = male], and pubertal status). Because the familial risk and assets indices included household sociodemographic measures (eg, poverty status), these measures were not used as covariates.

Sample descriptives are presented in Table 2. As shown in Table 3, familial risks and assets tracked well over time. Familial assets and child behavioral regulation were correlated in a positive direction, and both were inversely associated with familial risk.

Adjusted odds ratios (AORs) for all 5 cumulative indices predicting BMI trajectory membership are listed in Table 4, with nonoverweight (40th percentile) as the reference group. Results were similar when nonoverweight (70th percentile) served as the reference group and, thus, are not reported. Because we were interested in identifying factors that distinguish children who become severely obese from those who do not, we also present AORs with the overweight/obese trajectory as the reference group in Table 5. The results in Tables 4 and 5 are presented as bivariate models (evaluating independent effects of each predictor), multivariate models (presenting unique effects of each predictor in the presence of other predictors), and moderation models (presenting moderating effects of familial and child assets on the link between familial risks and BMI trajectories).

As shown in Table 4, children exposed to greater familial risk during 24 to 54 months had an increased odds of being in the nonoverweight (70th percentile) trajectory compared with the nonoverweight (40th percentile) trajectory. This effect became nonsignificant after familial assets and child behavioral regulation were added in subsequent models.

As shown in Table 4 (bivariate models), greater exposure to familial assets during both developmental periods and higher child behavioral regulation decreased the odds of being in the overweight/obese trajectory, whereas familial risk increased the odds relative to the nonoverweight (40th percentile) trajectory. Only familial assets (24–54 months) continued to remain statistically significant after both familial asset indices were entered into the same model (model 1) and after child behavioral regulation and familial risk (24–54 months) were added in model 4.

Among the moderation models in Table 4, the interaction term for familial risk by child behavioral regulation (24–54 months; model 7) predicted membership in the overweight/obese trajectory. As shown, among children with low behavioral regulation, greater exposure to familial risks (24–54 months) predicted membership in the overweight/obese group, whereas no relation was observed among children with high behavioral regulation. This result remained unchanged after adjusting for familial risk (model 8). It is noteworthy that a similar moderation effect was observed for the nonoverweight (70th percentile) BMI trajectory, despite the interaction term not reaching statistical significance for this group (simple effects are shown in Table 4).

As shown in Tables 4 and 5 (bivariate models), greater exposure to familial assets during both developmental periods and higher child behavioral regulation decreased the odds of being in the severely obese trajectory compared with the nonoverweight (40th percentile) and overweight/obese trajectories. No effects were observed for familial risk indices. Both familial assets (24–54 months) and child behavioral regulation continued to remain statistically significant, with nonoverweight (40th percentile) as the reference group, in multivariate models 1, 3, and 4 (Table 4).

As shown in Tables 4 and 5, the interaction terms for familial risk by familial assets (both developmental periods) were significant in their prediction of membership in the severely obese trajectory. Among children with higher familial assets (both periods), greater exposure to familial risks decreased the odds of membership in the severely obese group relative to the nonoverweight (40th percentile) and overweight/obese trajectories. In other words, higher familial assets buffered the negative effects of familial risks and decreased odds of membership in the severely obese group.

Finally, in Table 4, the interaction term for familial risk by child behavioral regulation (24–54 months) in model 7 predicted membership in the severely obese trajectory such that greater exposure to familial risks (24–54 months) increased the odds of membership in this group (P = .088) only among children with low relative behavioral regulation. No relation was observed among children with high behavioral regulation. This result did not change after adjusting for familial risk (model 8).

We took a novel, strength-based approach in identifying the protective effects of early exposure to familial psychosocial assets in infancy/toddlerhood (1–15 months) and early childhood (24–54 months) and child behavioral regulation on BMI growth from 2 to 15 years.³⁶  In our predominantly non-Hispanic, White, English-speaking sample, we found that exposure to familial assets (eg, maternal sensitivity) during early childhood outweighed the effects of exposure to the same set of familial assets during infancy, two sensitive periods of development. Exposure to familial assets in early childhood also reduced the likelihood of obesity and severe obesity, outweighed the positive effects of child behavioral regulation on obesity, and buffered the negative effects of concurrent familial risk on the development of severe obesity. Familial psychosocial assets, such as positive parenting styles and enriched home environments, have been linked to positive developmental outcomes (eg, language development, academic achievement)^37–39  and reduced obesity risk, and been shown to facilitate the development of behavioral regulation in childhood,⁴⁰  even among children who are socioeconomically disadvantaged.^41,42  Taken together, our findings highlight the potentially unique and salient role of familial psychosocial assets in promoting obesity resilience.

In the current study, child behavioral regulation reduced the risk for obesity and severe obesity and buffered the negative effects of familial risk among children having an overweight/obese trajectory. Early childhood is a critical period for the development of behavioral regulation processes, such as executive function (a collection of integrated neurocognitive processes involved in higher-order cognition).^43,44  Executive function has been highlighted as a potentially modifiable youth-level asset and linked to healthy eating behaviors, less sedentary behavior, greater physical activity,^7–9  and obesity resilience in childhood.^16,26,45 

It is noteworthy that children in the severely obese trajectory had lower family assets and behavioral regulation but did not have greater familial risk exposure compared with nonoverweight children. This finding is consistent with previous work in a socioeconomically disadvantaged sample wherein cumulative psychosocial risk did not distinguish children exhibiting a severely obese trajectory very early in childhood from those with a nonoverweight trajectory, although individual differences were observed in behavioral regulation.⁴⁵  Other unmeasured genetic and environmental variables may be more potent indicators of risk among children with severe obesity very early in life.

We used a strength-based approach by evaluating the protective role of familial psychosocial assets and child behavioral regulation on BMI growth trajectories. Because familial risk was included in the models, we were able to evaluate the independent and buffering role of family and child-based protective factors on BMI trajectories. A growing number of studies have focused on identifying factors of resilience that help to maintain nonobesity across the lifespan, despite the presence of obesity-related risk factors.^24,46–49  Resilience-based frameworks offer a solution-based approach to the study of obesity among at-risk populations^50–53  by leveraging assets already present among individuals and families in at-risk settings as well as by highlighting modifiable characteristics of resilient individuals and families.

Study limitations should be noted. A number of risks and assets were not addressed in the current study, including those within the home and neighborhood environments such as parental weight status, youth dietary and physical activity patterns, food insecurity, and social support, all of which may be important indicators of familial and child assets and potential correlates of weight gain during infancy and childhood.^54–59  Although the NICHD SECCYD sample largely mirrored the racial/ethnic composition of the United States on the basis of the 1990 census, the sample was predominantly non-Hispanic, White, not poor and excluded non-English speakers. Our findings cannot be generalized to all US families and, in particular, to children from ethnic and socioeconomically disadvantaged backgrounds. There is evidence to suggest that behavioral regulation may not confer the same degree of protection from obesity among Black children compared with White children.^60,61  Also unclear is whether the benefits of familial and child psychosocial assets persist as youth exposure to risk increases. Data collection for this sample occurred between 1991 and 2007, >14 years before this publication. Rates of obesity have increased since then, particularly among children who are ethnic minorities and poor.⁶²  Given the small sample sizes in some BMI trajectories, we were unable to explore differences by sex, race, or ethnicity. Finally, we did not address the stability and chronicity of risk and assets, given that data were not consistently available for all variables of interest.

The American Heart Association recommends that researchers identify resilience factors that buffer the negative effects of childhood adversity on later cardiometabolic outcomes.²  We found in the current study that family psychosocial assets and child behavioral regulation reduce the risk for obesity and severe obesity in youth and buffer the negative effects of early adversity on weight gain. Programs targeting these factors in early childhood have shown promise for obesity prevention.^63–69  Early childhood is a sensitive period for the development of self-regulation and positive parenting strategies, and obesity and severe obesity track strongly from childhood into adolescence and adulthood^70–72 ; thus, efforts to build strengths in these areas should occur as early in childhood as possible. Pediatricians and other clinicians who monitor children’s growth can use the present findings to tailor messages to parents about the potential value of improving familial and child psychosocial assets for the prevention of childhood obesity, particularly among children exhibiting early signs of an obesity-prone trajectory.

The Study of Early Child Care and Youth Development (SECCYD) Early Child Care Research Network was conducted by the Eunice Kennedy Shriver National Institute of Child Health and Human Development through a cooperative agreement that calls for scientific collaboration between grantees and the NICHD staff (U01 HD019897). The Pennsylvania State University have restricted data-use agreements to analyze the SECCYD data. The authors declare no conflicts of interest.

AOR

adjusted odds ratio

NICHD SECCYD

National Institute of Child Health and Human Development Study of Early Child Care and Youth Development