Low-Income Preschoolers with Higher Temperamental Surgency Enjoy and Respond More to Food, Mediating the Path to Higher Body Mass Index

Abstract

Background

Temperament is associated with obesity risk. However, the mechanisms linking temperament and eating behavior to childhood adiposity are unclear.

Objectives

To examine whether 3 temperament dimensions (surgency, effortful control, and negative lability) are uniquely associated with an increased BMI z score (BMIz) concurrently and an excessive rate of change in BMIz longitudinally through 4 eating behaviors (food responsiveness, enjoyment of food, emotional overeating, and satiety responsiveness) among low-income preschoolers, independent of home environment quality.

Methods

379 preschoolers were recruited from Head Start in the Midwest region of the United States. Primary caregivers reported child temperament, eating behaviors, and the level of chaos at home. Child BMIz was derived from weight and height measurements at ages 4, 5, and 6 years on average.

Results

Path analyses revealed that higher levels of surgency predicted more food responsiveness and enjoyment of food, which was in turn associated with higher concurrent BMIz, independent of effortful control, negative lability, and home chaos.

Conclusion

Low-income surgent preschoolers were more likely to have elevated BMIz as they were more inclined to eat in response to external cues and have a high appetitive drive. Obesity prevention programs might target low-income children with surgent temperaments, and the identified eating behaviors.

INTRODUCTION

Childhood obesity is common¹ and interventions are needed. Differences in temperament may explain why some young children are prone to obesity. Three temperament dimensions: surgency (high impulsivity, pleasure- and novelty-seeking, and activity level), effortful control (ability to refrain from a behavior, maintain attention, and resist distraction), and negativity (high lability, reactivity, and negative emotion)² have been linked to risk of excess adiposity in children. Specifically, greater surgency,³ impulsivity,⁴ hyperactivity,⁴ poor inhibitory control,^5,6 limited ability to delay gratification,^6,7 short attention span,⁸ and negative emotion^4,5,9 have been associated with accelerated weight gain and/or higher weight status.

The mechanisms linking these temperament features to obesity risk are not entirely clear. Differences in eating behavior such as external eating,^10–12 frequent desire to eat,^12,13 emotional overeating,^11,12 and under-responsiveness to internal satiety cues¹² have been implicated in the development of childhood obesity. Findings from studies that have examined associations between temperament and obesity-promoting eating behavior are not always in the expected direction. Specifically, children with lower inhibitory control are more vulnerable to external or emotional overeating.^14–17 Contrary to what might be predicted, however, children with more negative affectivity are more responsive to satiety and less interested in eating.^18,19 We have previously examined associations between dimensions of temperament and children’s eating behaviors,²⁰ but whether these temperament features predict body mass index (BMI), and how those associations may be explained by the child’s eating behaviors, remain unexamined.

Understanding the mechanisms linking temperament to elevated BMI is essential to developing effective interventions that can be tailored to children with the identified temperament characteristics. The present study sought to fill the following gaps in the present literature. First, no study has examined the mediating mechanisms linking these three temperament dimensions to adiposity through obesity-promoting eating behaviors. Second, children who experience early adiposity rebound (i.e., when BMI begins to increase from its nadir) have been reported to have an increased risk of obesity later in life.^21,22 Thus, preschool is a critical developmental period for examining factors that may contribute to obesity-promoting eating behaviors and elevated BMI.²³ Few studies have examined these associations in the preschool age range.^7,14 The present study, therefore, tested the hypothesis that among preschoolers the temperament dimensions of higher surgency, lower effortful control, and higher negative lability are each associated with greater adiposity and greater rates of increase in adiposity, and that these associations are mediated through greater obesity-promoting eating behaviors (food responsiveness, enjoyment of food, emotional overeating, and satiety responsiveness). Furthermore, given that preschoolers have been shown to be more likely to engage in obesogenic eating or at greater risk for obesity when living in a chaotic home environment,^20,24 home environment quality (operationalized as level of chaos in the home) was considered as a covariate.

METHODS

Study Design and Participants

Participants were preschool-age children from the Midwest region of the United States who were enrolled in a longitudinal study of child eating behavior, with three assessments at age 4, 5, and 6 years on average. Families were recruited from Head Start, a federally funded preschool program for low-income children. Inclusion criteria were that the child was born at 35 weeks gestation or more and did not have any significant perinatal or neonatal complications, developmental disabilities, medical problems, or food allergies; child was not in foster care; caregiver and child were English-speaking; and caregiver did not have a college degree. Trained research assistants administered questionnaires orally to caregivers at the first time-point assessment. The study was approved by the University of Michigan Institutional Review Board. The current sample includes 379 participants with complete data for all variables in the analyses.

Measures

Predictors: Child Temperament

Children’s primary caregivers completed the Children’s Behavior Questionnaire (CBQ),² on which they rated how well each item describes the child (1 = extremely untrue; 7 = extremely true). The Surgency dimension is the mean of 25 items (e.g., “often rushes into new situations”; α = .70), capturing children’s activity level, high intensity pleasure, impulsivity, and shyness (reversed). The Effortful Control dimension was measured using the mean of 12 items (e.g., “can easily stop an activity when she/he is told “no””; α = .76), capturing children’s attention focusing and inhibitory control. Only the CBQ Surgency and Effortful Control scales were used due to time constraints. Caregivers also completed the 16-item Negative Lability scale of the Emotion Regulation Checklist (ERC).²⁵ Items (e.g., “is easily frustrated”) were rated on a 1 (rarely/never) to 4 (almost always) point scale and a mean was calculated (α = .85). On each of the scales, higher scores represent higher levels of the temperament dimension.

Mediators: Child Eating Behavior

Caregivers completed the Children’s Eating Behavior Questionnaire (CEBQ),¹³ a 35-item questionnaire to which caregivers responded on a scale of 1 (never) to 5 (always). Subscale scores are calculated as the mean of the contributing items, such that higher scores represent more of the given behavior. Subscales included in this analysis were: food responsiveness assessing food consumption in response to external cues (e.g., “my child is always asking for food”; 5 items; α = .84); enjoyment of food capturing appetitive drive, desire to eat, and interest in eating (e.g., “my child loves food”; 4 items; α = .84); emotional overeating reflecting a tendency to seek comfort through eating in response to emotional arousal (e.g., “my child eats more when anxious”; 4 items; α = .79); and satiety responsiveness referring to the ability to recognize internal satiety cues (e.g., “my child gets full easily”; 5 items; α = .73).

Outcomes: Concurrent BMIz and Rate of Change in BMIz

Children were weighed and measured without shoes or heavy clothing by trained research assistants according to standard protocols, using a ±0.1 kg calibrated scale and a ±0.1 cm calibrated stadiometer. Measurements were collected again an average of 0.7 (SD = 0.5) and 1.7 (SD = 0.5) years later. BMI was calculated and BMI z score derived using age- and sex-specific growth reference charts from the US Centers for Disease Control and Prevention.²⁶

Covariate: Home Environment Quality

Given reported associations of home environment quality with temperament,¹⁴ eating behaviors,^{14 Error! Bookmark not defined}. and obesity,²⁴ home environment quality was examined as a covariate. The Confusion, Hubbub, and Order Scale (CHAOS)²⁷ was used to capture the extent to which the child’s home environment is characterized by disorganization, confusion, noise, and a lack of routine (i.e., chaos). Caregivers indicated whether each of the 15 items was true or false and following appropriate reverse coding. Items were summed (α = .80) such that higher scores represent greater home chaos.

Statistical Analysis

Hierarchical linear models (HLM)²⁸ using random parameters were used to capture individual BMIz growth curves for each participant. This approach accounts for the time differential in the measurement of the BMIz in a direct way using the parametric function of the rate of change in BMIz per year. The random intercept is an estimate of the expected BMIz at the age of 4 years for a given individual, and the random slope is the expected rate of change in BMIz from ages 3 to 8 years.

Multivariate analyses accounting for inter-correlations among independent variables were implemented using path analyses to examine whether eating behaviors mediated the effect of child temperament on BMIz growth pattern (both intercept and slope), independent of home chaos. Four separate path models were run testing each of the 4 eating behaviors (food responsiveness, enjoyment of food, emotional overeating, and satiety responsiveness), with the 3 temperament dimensions (surgency, effortful control, and negative lability) and home chaos included (see Figure 1). The covariance path between surgency and home chaos was non-significant and therefore excluded from the path analyses to obtain a more parsimonious model with a better fit. All analyses were conducted in AMOS 21 using the maximum-likelihood estimation procedure. The nonparametric bootstrapping procedure was used for inferences on the indirect effects. The model fit was assessed based on the χ² goodness-of-fit statistic being non-significant with a p-value > .05, comparative fit index (CFI) > .90, and standardized root mean residual (SRMR) < .06.29

Figure 1. Path Model for the Links between Temperament, Eating Behavior, and BMIz Growth Pattern, Controlling for Home Chaos.

Note. Four eating behaviors including food responsiveness, enjoyment of food, emotional overeating, and satiety responsiveness were tested individually in each of four models. Inter-correlations between exogenous variables were estimated in each model; the double arrows and correlation estimates were not included in the figure above to make the figure as parsimonious as possible.

† p < .10. ^* p < .05. ^** p < .01. ^*** p < .001. All models have good fit with model-fit statistics on the recommended range: For each model, the χ²-fit statistics was highly non-significant p-value > .05, CFI > .9, and SRMR <. 06.

RESULTS

Demographic characteristics of the sample are presented in Table 1. Child mean age at the initial assessment was 4.2 years. Males and females were equally represented. Standardized path coefficients for the four model are presented in Figure 1. All 4 models fit the data well, with p-values for the χ² goodness-of-fit statistic ranging from .347 to .665, CFI ranging from .997 to 1.00, and SRMS ranging from .029 to .32.

Table 1.

Prevalence of overweight and child care use among 3– and 5–year–old children in the LATE project.

		3–year–olds	5–year–olds	total	p–value*
	N	781	902	1683
Overweight %		10	14	12	*
Current type of childcare %	parental	43	27	34
	Informal	3	2	2
	(group) family childcare	20	15	18
	Childcare centre	34	57	46	***
Number of childcare places %	0	28	11	19
	1	51	40	45
	2	16	29	23
	3 or more	5	20	13	***
Age when entering childcare %	under 1 year	11	10	11
	1.00–1.99 years	38	35	36
	2.00–2.99 years	19	20	19
	at least 3	2	23	13
	Has not entered	31	12	21	***
Hours spent in childcare per day %	0	36	21	28
	max 5 hours	6	7	6
	5.1–7.9 hours	21	25	23
	8 hours	27	33	30
	more than 8 hours	11	15	13	***
Days in childcare per month %	0	36	22	29
	under 20	24	30	27
	20 or more	40	48	44	***
Total childcare time per month %	0	37	23	29
	under 120 hours	19	23	21
	120–159 hours	16	21	19
	160 h or more	28	33	31	***

*for difference between age groups: * <0.05, **<0.01, *<0.001

Food responsiveness significantly mediated the association between surgency and BMIz. Higher surgency was associated with more food responsiveness, β = .14. More food responsiveness in turn was associated with higher BMIz, β = .21. The indirect effect of surgency on BMIz through food responsiveness was significant. Food responsiveness did not mediate the association between surgency and the rate of increase in BMIz. The BMIz growth pattern at high versus low levels of food responsiveness (defined as 1 standard deviation above and below the mean) is shown in Figure 2a.

Figure 2.

BMIz Growth Pattern at High versus Low Levels of Eating Behaviors

Enjoyment of food significantly mediated the association between surgency and BMIz. Higher surgency was associated with more enjoyment of food, β = .23. More enjoyment of food in turn was associated with higher BMIz, β = .27. The indirect effect of surgency on BMIz through enjoyment of food was significant. Enjoyment of food did not mediate the association between surgency and the rate of increase in BMIz. The BMIz growth pattern at high versus low levels of enjoyment of food (defined as 1 standard deviation above and below the mean) is shown in Figure 2b.

Neither effortful control nor negative lability was significantly associated with eating behaviors, BMIz intercept, or BMIz slope. Given these non-significant paths, there was no evidence for an indirect effect of effortful control or negative lability on BMIz intercept or slope through eating behaviors. The four mediation models were also tested in boys and girls separately and no substantial differences in the pattern of results were found between boys and girls.

DISCUSSION

This study had 3 main findings. First, more surgent preschoolers were more responsive to food cues in the environment and enjoyed food more, each of which in turn was associated with higher BMIz. Each of these two eating behaviors fully mediated the association of surgency with BMIz. These findings suggest that associations between surgent temperament and obesity risk are operating through eating behavior among low-income preschoolers. Obesity prevention programs may thus increase their effectiveness by focusing on low-income young children with high surgency who are particularly at risk, and attempting to modify the food environment for these children. Specifically, our findings suggest that because surgent children particularly enjoy food and attend to food cues in the environment, reducing their exposure to the pervasive food cues in the current environment may be especially important for this subgroup.

Second, contrary to our hypothesis, greater surgency was not associated with greater increases in low-income children’s BMIz from ages 4 to 6 years. Rather, children described by their caregivers at age 3–4 years as having high surgency, food responsiveness, and enjoyment of food already had a higher BMI relative to their same-age peers and that higher BMI continued to track at the higher level for the nearly 2 year follow up period. This observation suggests that the links between a surgent temperament and these obesity-promoting eating behaviors may need to be addressed prior to age 3 years to effectively prevent unhealthy increases in BMI.

Third, neither effortful control nor negative lability was associated with eating behaviors or BMI. Our study is not the first to be unable to detect a significant association of effortful control with BMI in preschool-aged children.^6,7 Links between effortful control and BMI may emerge only later in childhood, when self-regulation specific to eating (e.g., self-restraint in response to palatable food) may become a more expected behavior. Indeed, at least one study has suggested that links between self-regulatory capacity and BMI only emerge later in childhood.⁶ Mechanisms by which effortful control may be associated with BMI longitudinally deserve further consideration in future research. Although there has been interest in negativity, sometimes characterized as “difficult temperament”, as a predictor of obesity risk in children,^4,5 the literature suggesting that negative affectivity is actually associated with undereating or selective eating behaviors^18–20 has contributed to uncertainty about this association. In this study, negative lability primarily captured anger and distress to limits, which might contribute to the inability to detect an association with either eating behavior or BMI. Future work might consider measuring temperamental negativity and moderators of the association such as child age, socioeconomic status, or parenting that could be masking a direct association.

Strengths of the present study include using repeated objective measures of height and weight, reliable and valid questionnaires, and HLM to capture individual BMIz growth curves. Nevertheless, findings should be interpreted in light of the study limitations. This study focused on temperamental negative lability and did not capture all types of temperamental negativity. In addition, the results may not be generalizable to populations that are not low-income preschoolers from the Midwestern United States.

In summary, this study found that low-income preschoolers with a surgent temperament are more likely to respond to food cues in the environment and to enjoy food, and these behaviors, in turn, are associated with greater adiposity by age 3–4 years. Clinicians might consider explaining the risks of this behavior pattern to caregivers in infancy and toddlerhood, so that caregivers can be attentive to the food environment and how they manage the child’s eating behavior.

Table 2.

Odds ratio (OR) and 95% confidence intervals (CI) for overweight according to child care use in the LATE–project. Logistic regression.

		model 1		model 2
		OR (95 % CI)	p	OR (95 % CI)	p
Current type of childcare	parental care	1		1
	informal care	1.70 (0.71–4.04)		1.88 (0.65–5.45)
	(group) family childcare	1.34 (0.88–2.03)		1.59 (0.98–2.58)
	childcare centre	0.97 (0.69–1.37)		1.10 (0.73–1.61)
			0.26		0.20
Number of childcare places*	total	1.14 (0.99–1.30)	0.07	1.14 (0.99–1.32)	0.08
	girls	1.29 (1.11–1.51)	0.001	1.33 (1.11–1.60)	0.003
	boys	0.96 (0.76–1.20)	0.70	0.84 (0.65–1.10)	0.21
Age when entering childcare	at least 3	1		1
	2.00–2.99 years	1.39 (0.80–2.44)		1.89 (1.00–3.60
	1.00–1.99 years	1.22 (0.73–2.05)		1.41 (0.77–2.58)
	under 1 year	2.53 (1.41–4.52)		3.13 (1.62–6.03)
	Has not entered	1.30 (0.75–2.27)		1.37 (0.70–2.66)
			0.01		0.003
Hours in childcare per day	0	1		1
	max 5 hours	0.63 (0.30–1.33)		0.93 (0.42–2.03)
	5.1–7.9 hours	0.90 (0.59–1.38)		1.34 (0.82–2.18)
	8 hours	0.88 (0.59–1.31)		1.17 (0.73–1.87)
	more than 8 hours	1.28 (0.81–2.03)		1.51 (0.88–2.59)
			0.34		0.54
Days in childcare per month	0	1		1
	under 20	0.88 (0.59–1.32)		1.23 (0.78–1.93)
	at least 20	0.97 (0.68–1.38)		1.17 (0.77–1.79)
			0.82		0.65
Total childcare time per month	0	1		1
	under 120 hours	0.90 (0.59–1.39)		1.32 (0.82–2.13)
	120–159 hours	0.99 (0.64–1.54)		1.36 (0.82–2.25)
	160 h or more	0.99 (0.68–1.45)		1.11 (0.70–1.76)
			0.97		0.57

Each childcare attendance variable was examined separately of each other.

model 1: Adjusted with gender and age.

model 2: Adjusted with model 1 + mother's BMI, father’s BMI, highest education in the family, family structure, smoking during pregnancy, birth weight.

^*Continuous variable. Analyses were also done separately for boys and girls, because gender interaction was found.

Table 3.

Odds ratio (OR) and 95% confidence intervals (CI) for overweight according to childcare use in the LATE–project. Logistic regression. 3- and 5-year-olds separately.

		3-year-olds				5-year-olds
		model 1		model 2		model 1		model 2
		OR (95 % CI)	p	OR (95 % CI)	p	OR (95 % CI)	p	OR (95 % CI)	p
Current type of childcare	parental care	1		1		1		1
	informal care	2.04 (0.65–6.43)		1.71 (0.41–7,19)		1.40 (0.37–5.29)		2.09 (0,41– 10.67)
	(group) family childcare	1.38 (0.76–2.51)		1.80 (0.91–3,55)		1.31 (0.73–2.35)		1.48 (0.72–3.02)
	childcare centre	0.88 (0.50–1.57)		1.27 (0.67–2,43)		0.87 (0.56–1.37)		1.00 (0.58–1.73)
			0.35		0.39		0.45		0.52
Number of childcare places*	total	1.02 (0.78–1.34)	0.88	1.13 (0.86–1.48)	0.39	1.17 (1.00–1.37)	0.05	1.13 (0.94–1.35)	0.20
	girls	1.07 (0.77–1.50)	0.67	1.22 (0.87–1.73)	0.26	1.33 (1.10–1.61)	0.003	1.38 (1.09–1.74)	0.01
	boys	0.94 (0.60–1.46)	0.78	0.97 (0.59–1.58)	0.90	0.90 (0.67–1.20)	0.46	0.78 (0.56–1.09)	0.15
Age when entering childcare	at least 3	1		1		1		1
	2.00–2.99 years	0.64 (0.17–2.44)		0.72 (0.16–3.23)		1.56 (0.80–3.02)		1.85 (0.86–3.96)
	1.00–1.99 years	0.37 (0.10–1.39)		0.49 (0.12–2.08)		1.83 (1.02–3.27)		1.54 (0.78–3.05)
	under 1 year	0.92 (0.23–3.66)		1.17 (0.26–5.33)		3.38 (1.71–6.71)		3.88 (1.80–8.36)
	Has not entered	0.48 (0.13–1.80)		0.45 (0.11–1.95)		2.13 (1.06–4.25)		1.77 (0.77–4.09)
			0.09		0.14		0.01		0.01
Hours in childcare per day	0	1		1		1		1
	max 5 hours	1.20 (0.44–3.31)		1.42 (0.47–4.34)		0.37 (0.12–1.09)		0.62 (0.20–1.95)
	5.1–7.9 hours	0.93 (0.48–1.80)		1.40 (0.66–2.97)		0.85 (0.48–1.48)		1.14 (0.59–2.22)
	8 hours	1.13 (0.63–2.03)		1.88 (0.95–3.74)		0.72 (0.42–1.24)		0.82 (0.42–1.57)
	more than 8 hours	0.97 (0.42–2.22)		1.60 (0.64–3.99)		1.34 (0.75–2.38)		1.31 (0.64–2.65)
			0.98		0.49		0.09		0.54
Days in childcare per month	0	1		1		1		1
	under 20	1.18 (0.64–2.16)		1.50 (0.76–2.98)		0.71 (0.42–1.20)		1.00 (0.54–1.86)
	at least 20	1.01 (0.45–1.15)		1.59 (0.84–2.99)		0.90 (0.56–1.44)		0.88 (0.49–1.57)
			0.84		0.32		0.43		0.86
Total childcare time per month	0	1		1		1		1
	under 120 hours	1.33 (0.71–2.49)		1.75 (0.86–3.57)		0.67 (0.37–1.21)		1.01 (0.52–1.95)
	120–159 hours	0.97 (0.48–1.98)		1.35 (0.60–3.04)		0.96 (0.55–1.67)		1.18 (0.60–2.31)
	160 h or more	1.03 (0.56–1.86)		1.77 (0.88–3.55)		0.93 (0.56–1.53)		0.80 (0.43–1.49)
			0.81		0.33		0.57		0.66

Each childcare attendance variable was examined separately of each other.

model 1: Adjusted with gender and age.

model 2: Adjusted with model 1 + mother's BMI, father’s BMI, highest education in the family, family structure, smoking during pregnancy, birth weight.

^*Continuous variable. Analyses were also done separately for boys and girls, because gender interaction was found.

What is already known about this subject?

• Three temperament dimensions (surgency, effortful control, and negative lability) have been associated with the risk of excess adiposity in children.

• However, the associations between temperament dimensions and obesity-promoting eating behaviors are not always in the expected direction, and the mechanisms linking temperament dimensions to obesity risk are not entirely clear.

What this study adds?

• This study examined the mediating mechanisms linking three temperament dimensions to adiposity through obesity-promoting eating behaviors among low-income preschoolers.

• Results showed that more surgent preschoolers enjoyed and responded more to food; these behaviors, in turn, were associated with elevated BMI.

Abbreviations

BMIz

body mass index z score

CBQ

Children’s Behavior Questionnaire

ERC

Emotion Regulation Checklist

CEBQ

Children’s Eating Behavior Questionnaire

CHAOS

Confusion, Hubbub, and Order Scale

HLM

Hierarchical linear models

CFI

comparative fit index

SRMR

standardized root mean residual