Child eating behaviors are consistently linked to intake across meals that vary in portion size

Abstract

Prior studies evaluating a single meal in children characterized an “obesogenic” style of eating marked by larger bites and faster eating. It is unclear if this style is consistent across portion sizes within children so we examined eating behaviors in 91 children (7-8 years, 45 F) without obesity (BMI<90^th percentile). Children consumed 4 ad libitum meals in the laboratory consisting of chicken nuggets, macaroni, grapes, and broccoli that varied in portion size (100%, 133%, 166%, 200%) with a maximum of 30 min allotted per meal. Anthropometrics were assessed using age and sex adjusted body mass index (BMI) percentile and dual energy x-ray absorptiometry. Bites, sips, active eating time, and meal duration were coded from meal videos; bite size (kcal and g/bite), proportion of active eating (active eating time/meal duration), and eating rate (kcal and g/meal duration) were computed. Intraclass correlation coefficients (ICC) showed that most eating behaviors were moderately consistent across portions (>0.50). The consistency of associations between eating behaviors and total meal intake and adiposity was assessed with general linear models adjusted for food liking, pre-meal fullness, age, and sex. Across all portions, more bites, faster eating rate, and longer meal duration were associated with greater intake. While higher BMI percentile was associated with faster eating rates across all meals, greater fat mass index was only associated with faster eating at meals with portions typical for children (i.e., 100% and 133%). In a primarily healthy weight sample, an ‘obesogenic’ style of eating was a consistent predictor of greater intake across meals that varied in portion size. The consistent relationship of these behaviors to intake makes them promising targets to reduce overconsumption.

1. Introduction

While what children eat is well-recognized as a contributor to pediatric obesity (J. Kim & Lim, 2019; Neri et al., 2022; Nguyen et al., 2023; Pérez-Escamilla et al., 2012), there is less known about the contribution of how children eat as it relates to positive energy balance. Recent studies focused on characterizing how children eat have identified eating styles based on meal microstructure (e.g., larger bites, faster eating speed) that are associated with greater energy intake at a meal (Fogel et al., 2017a, 2017b). However, a critical question remains about whether this eating style is stable across meals, especially since it is known that eating behaviors are influenced by meal characteristics. Portion size, in particular, has a robust effect on the amount of food consumed (Reale et al., 2019; Rolls et al., 2002). Serving larger portions leads to greater intake across different types of food and age groups (Reale et al., 2019). In addition, there is evidence that larger portion sizes can influence aspects of meal microstructure by increasing bite size in children (Fisher, 2007; Fisher et al., 2003; Gómez-Zúñiga & Wintergerst, 2023); therefore, the relative stability of eating style across meals that vary in portion size requires investigation. The current paper examined eating styles (e.g., bite size, eating rate) across four meals that varied in portion size to determine whether these behaviors were stable across conditions where attributes that influence total energy intake (i.e., portion size) were changed.

Prior research in children has characterized styles of eating associated with greater food consumption and obesity (Fogel et al., 2017a; Pearce et al., 2022). Eating rate, in particular, has been shown to be heritable (Epstein et al., 1976; Llewellyn et al., 2008) and positively associated with energy intake (Epstein et al., 1976; Fogel et al., 2017a, 2017b; Llewellyn et al., 2008) and obesity (Barkeling et al., 1991; Fogel et al., 2017a, 2017b; Geller et al., 1981; Llewellyn et al., 2008) in children. Additionally, faster bite (e.g., bites/min) and eating rates (e.g., kcal/min) and shorter meal duration at 4 years of age were associated with greater odds of having overweight (BMI>85th percentile) at 6 years of age (Berkowitz et al., 2010). While pilot interventions have aimed to reduce eating speed in children, there is limited evidence that focusing solely on eating rate is sufficient to produce changes in intake or adiposity (Cox et al., 2022). While, fewer studies have characterized the cumulative contribution of multiple eating behaviors (e.g., bite size, bite number, eating rate) to overall energy intake (Pearce et al., 2022), work from the GUSTO cohort suggests that larger bites, reduced oral processing per bite, and faster eating rates were all associated with greater intake in children (Fogel et al., 2017b, 2017a). Together, faster eating, reduced oral processing time, and larger bite size have been characterized as an ‘obesogenic’ style of eating associated with increased food intake and weight status. If this style is a stable predictor of excess energy intake across eating events, it could hold promise as a target for treating pediatric obesity.

Given the increased availability of large portions of palatable foods targeted to children in the marketplace (Young & Nestle, 2021), it is important to determine whether children’s oral processing behaviors are stable across meals that vary in the amount of food served. Interventions aiming to reduce excess food intake will likely be more effective if they target behaviors that are consistent across eating contexts and environments. To strengthen the evidence that meal microstructure could be a promising target for interventions to reduce excess weight, the current study aimed to establish the stability of children’s eating behavior across four laboratory ad libitum meals that served the same foods, but varied in amount of food served, or portion size. Additionally, the study aimed to extend findings from Fogel and colleagues (2017a) by testing whether the previously defined “obesogenic” style of eating would consistently predict increased food intake across meals that varied in portion size (Fogel et al., 2017a). In addition to increased food intake, we also examined whether this style of eating would be positively associated with child adiposity. It was hypothesized that children would show consistent patterns of eating behavior across portion size and that an ‘obesogenic style’ of eating characterized by a greater number of bites, larger bites, and faster eating would be associated with excess energy intake and adiposity, regardless of the amount of food served.

2. Methods

2.1. Participants

Children were enrolled in a longitudinal, 7-visit study (6 baseline visits and 1 follow-up visit a year later; supplemental Figure S1) aimed at identifying neurocognitive contributors to obesity (ClinicalTrials.gov NCT03341247). Of the 93 children enrolled, 91 children (46 male) aged 7-8 years old (M = 7.8, SD = 0.6) had usable videos of at least two baseline meals. The number of children who had usable video data at each meal ranged from 83-89 and 78 children had complete behavior data coded across all 4 meals. By design, children did not have obesity (body mass index < 90^th percentile). The sample was limited in ethnic and racial diversity (Table 1), but was reflective of the community in central Pennsylvania. Per inclusion criteria, children were in good health based on parent self-report with no learning or neurodevelopmental disabilities (e.g., ADHD, dyslexia) and not taking medications known to influence appetite or cognition. Children were excluded if they were colorblind, could not read at grade level, were not fluent in English, or had MRI contraindications (e.g., metal in the body; fMRI data reported separately). As part of primary study aims, children were also excluded if the biological mother did not meet criteria for one of the familial risk groups (high-risk: maternal BMI ≥ 30 kg/m²; low-risk: maternal BMIs ≤ 25 kg/m²). Child assent and parental consent were obtained in accordance with the Institutional Review Board of The Pennsylvania State University and families were compensated for each study visit completed (STUDY00020463).

Table 1.

Participant Characteristics

	Full Sample N = 91	Males N = 46 (51%)	Females N = 45 (49%)
	Mean (SD)	Mean (SD)	Mean (SD)
Age, yr	7.8 (0.6)	7.7 (0.7)	7.9 (0.6)
BMI percentile	47.8 (24.6)	48.3 (23.8)	47.3 (25.7)
Fat Mass Index, kg/m2	4.5 (0.09)	4.2 (0.8)	4.7 (1.0)
	N (%)	N (%)	N (%)
Race
White	88 (97%)	43 (93%)	45 (100%)
Asian	3 (3%)	3 (7%)	0 (0%)
Income
>$100,000	31 (38%)	13 (28%)	20 (44%)
$51,000 - $100,000	44 (50%)	26 (57%)	18 (40%)
<$51,000	11 (12%)	6 (13%)	5 (11%)
Unknown	3 (3%)	1 (2%)	2 (4%)
Maternal Education
> Bachelor’s Degree	28 (31%)	13 (28%)	15 (33%)
Bachelor’s Degree	44 (49%)	21 (46%)	23 (51%)
Associate/Technical Degree	9 (10%)	7 (15%)	2 (4%)
High School/GED	9 (10%)	4 (9%)	5 (11%)
Unknown	1 (1%)	1 (2%)	0 (0%)

GED: General Education Development certificate. Note: 100% of participants reported ethnicity of Non-Hispanic/Latinx. Percent values may not sum to 100% due to rounding.

2.2. Protocol

This study focused on baseline data related to anthropometrics (visit 1) and children’s eating behavior across four laboratory meals that varied in amount of food served (meals served on visits 2-5). Children received each of the four portion size meals and the order in which children received the meals was randomly assigned (using a Latin square design) and counterbalanced (Figure 1). Children fasted for 3 hours before each visit to achieve a typical pre-meal appetitive state. Prior to each meal, participants rated their fullness on a child-friendly visual analog scale (Keller et al., 2006) and then tasted and rated liking of small samples (~2-3 grams) of each food item using a 5-point Likert scale (1-‘Hate It’ to 5-‘Love It’).

Figure 1.

Pictures of the 4 portion size meals as served to participants.

2.2.1. Demographic Characteristics.

The parent that accompanied the child to the visit, which was the mother 87% of the time (n = 79), completed a demographic survey. Socioeconomic status was assessed through yearly family income and years of maternal education (Bradley & Corwyn, 2002).

2.2.2. Anthropometrics.

During visit 1, the accompanying parent and child had height and weight measured twice using a stadiometer and standard scale (Scale_Tronix Model 5002). Body mass index (BMI; kg/m²) was calculated and child weight status was determined using BMI percentile according to the Centers for Disease Control and Prevention (Kuczmarski et al., 2002). Whole-body dual energy x-ray absorptiometry (DXA; Hologic Inc.) was used to measure children’s body composition. DXA uses low-energy x-rays to measure various aspects of body composition including amounts of lean tissue, fat mass, and bone density. Fat mass index (FMI), a measure of adiposity that accounts for height and has been advocated for use in pediatric longitudinal studies (Kakinami et al., 2014),was calculated as total fat mass (kg)/height (m²). FMI was chosen because it accounts for child height and is more strongly associated with adiposity (Vanderwall et al., 2017)) and more predictive of metabolic health (Liu et al., 2013) than BMI.

2.2.3. Meal Paradigm.

Meals included four commonly consumed foods that varied in energy density – macaroni and cheese (1.7 kcal/g), chicken nuggets (2.5 kcal/g), broccoli (1 kcal/g), and grapes (0.7 kcal/g; see Table S1 for brands and Figure 1 for meal photos). While the meal was not designed a priori to compare differences in microstructure by food texture, the foods did vary by texture and, hence, the amount of oral processing required (e.g., broccoli requires greater oral processing relative to macaroni and cheese). The amount of each food served increased by 33%, 66% and 99% relative to the baseline portion (+/− 3% for macaroni and cheese, broccoli, and grapes and +/− 1 piece for chicken nuggets; Table S1). To maintain meal portion sizes at each condition, no extra servings were provided. Full meal preparation protocols are available on Open Science Framework (OSF) along with the study Standard Operating Procedures manual (osf.io/3j6s2/). The same amount of water was served at each meal and consumed ad libitum. All meals were consumed at either lunch (11 am – 1 pm) or dinner times (5 pm – 7 pm) and occurred ~1 week apart with the order of the four portions counterbalanced across participants. Children were given 30 minutes to eat and were told that they could eat as much or as little as they wanted. During the meal, children were read a non-food related book to provide a neutral distraction and to avoid the awkwardness of having them eat alone (pre-COVID-19: researcher read a book; post-COVID-19: computer audiobook). Food intake (pre-weight minus post-weight) was weighed to the nearest 0.1 g. Food weight was converted to energy using food manufacturer information (i.e., nutrition facts panel) and reputable nutrition databases (e.g., USDA). Overall intake, fullness, and food liking are presented in supplemental Table S2.

2.2.4. Microstructure Coding.

To characterize children’s eating behaviors, meals were conspicuously video recorded using an Axis M3004-V network camera, which was placed in the upper corner of the room. Behaviors were coded using Noldus Observer XT v16 (Noldus, 1991) according to a coding manual (Pearce et al., 2023) developed based on recommended best practices (Hetherington & Rolls, 2018; Pearce et al., 2022). All meals were coded by two, independent, trained coders (see Coding Manual for training protocol). Inter-coder reliability was assessed using two-way mixed effects intra-class correlation coefficients (ICC) with absolute agreement for each coded behavior (ICC > 0.86; Table S3). Coded behaviors were used to compute bite size, bite/sip/eating rates, and percent active eating time (i.e., (active eating time, min)/(meal duration, min) x 100). All microstructure behaviors are defined in Table 2 and mean and standard deviations are reported across meals in Table 3.

Table 2.

Microstructure coding definitions

Coded Behaviors
Behavior	Definition	Notes/Examples
Start of Meal	Time at which meal is placed in front of child
End of Meal	Time at which: 1) the child pushes meal tray aside to indicate completion of eating/meal; 2) the researcher moves tray away from child after the 30 min limit; or 3) child indicates they are finished eating
Bite	An act which results in a piece of food being placed in the mouth, not spat out, and subsequently chewed and swallowed	Nibbling. gnawing, and licking were coded as a single bite if: 1) it resulted in the consumption of food and 2) the child did not clearly pull food away from mouth or pause
Sip	An act which results in water entering the mouth and being subsequently swallowed	Guzzling (i.e., taking multiple fast swallows) was coded as a single sip if: 1) it resulted in consumption of water and 2) the child did not clearly pull drink away or pause
Latency to First Bite	Duration between start of meal and the child’s first bite OR sip
Meal Duration	Duration between start and end of meal
Active Eating Time	Duration of time actively engaged in eating behaviors (chewing, biting, drinking) or preparatory behaviors (e.g., arranging food, holding fork, rearranging items, etc.). Preparatory behaviors were defined as meal- or food-related behaviors that occur within 15 seconds of an eating behavior.	Example: moving food with a fork is a preparatory behavior if it leads to a bite within 15 seconds but would be considered playing with food if no bite is taken within 15 seconds
Computed Behaviors
Behavior	Definition
Bite Size	Intake (g or kcal) per bite	Intake (g or kcal)/Bites (N)
Bite Rate	Number of bites per minute of meal duration	Bites (N)/Duration (min)
Active Bite Rate	Number of bites per minute of active eating time	Bites (N)/Active Eating Time (min)
Sip Rate	Number of sips per minute of meal duration	Sips (N)/Duration (min)
Active Sip Rate	Number of sips per minute of active eating time	Sips (N)/Active Eating Time (min)
Eating Rate	Intake (g or kcal) per minute of meal duration	Intake (g or kcal)/Duration (min)
Active Eating Rate	Intake (g or kal) per minute of active eating time	Intake (g or kcal)/Active Eating Time (min)
Percent Active Eating	Percent of meal duration that was included in active eating time	[Active Eating Time (min)/Meal Duration (min)] X 100

Table 3.

Meal Microstructure Across Meals

	Portion 1	Portion 2	Portion 3	Portion 4	ICC
	N = 89	N = 88	N = 87	N = 83
	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)	N all meals	ICC
Bites, N	72.6 (37.9)	75.7 (37.8)	82.3 (45.1)	79.3 (44.0)	78	.70
Sips, N	6.6 (6.5)	7.9 (8.3)	6.9 (6.6)	5.6 (6.0)	78	.53
Latency to 1st Bite, min	0.4 (0.3)	0.4 (0.2)	0.4 (0.2)	0.4 (0.3)	78	.20
Meal Duration, min	18.0 (8.8)	18.5 (8.4)	19.4 (8.4)	19.2 (8.8)	78	.70
Active Eating Time, min	15.3 (7.3)	15.9 (7.0)	17.1 (7.6)	17.2 (8.0)	78	.68
Bite Rate, bites/min	4.3 (1.8)	4.3 (1.8)	4.5 (2.0)	4.4 (1.9)	78	.73
Active Bite Rate, bites/min (active)	4.9 (1.8)	4.9 (1.8)	5.0 (2.0)	4.8 (1.8)	78	.76
Sip Rate, sips/min	0.4 (0.3)	0.4 (0.5)	0.4 (0.4)	0.3 (0.3)	78	.52
Active Sip Rate, sips/min (active)	0.4 (0.4)	0.5 (0.5)	0.4 (0.4)	0.3 (0.4)	78	.52
Bite Size, g/bite	6.4 (2.4)	6.7 (2.8)	6.4 (2.2)	7.0 (5.2)	75	.32
Bite Size, kcal/bite	7.7 (3.3)	7.9 (3.1)	7.8 (3.2)	8.9 (6.5)	75	.39
Eating Rate, g/min	25.9 (11.3)	26.5 (10.5)	26.4 (10.9)	27.2 (12.5)	75	.66
Eating Rate, kcal/min	31.4 (14.8)	32.2 (14.7)	32.3 (16.0)	35.0 (16.4)	75	.67
Eating Rate, g/min (active)	29.7 (11.9)	30.3 (11.5)	29.6 (11.3)	30.5 (15.0)	75	.60
Eating Rate, kcal/min (active)	35.9 (16.1)	36.5 (15.7)	35.9 (16.6)	39.1 (19.5)	75	.63
Percent Active Eating, active/meal duration	87% (12%)	88% (11%)	89% (12%)	90% (11%)	78	.26

g: grams, ICC: intraclass correlation coefficient; kcal: kilocalories, min: minutes, SD: standard deviation

Note: A total of 91 children had data for 2 or more meals, however, the sample size for each meal is lower because different children missed different meals due to counterbalancing the order of meals and random technical errors in video recording.

2.3. Analytic Approach

2.3.1. Analyses were conducted in R and all code and data are publicly available (osf.io/3j6s2/; https://github.com/alainapearce/FBS_OBstyle_microstructure).

The effect of portion size served on child intake from this cohort has been reported elsewhere (Keller et al., 2023) and therefore, these results will not be reported in the current paper.

2.3.2. Descriptive Analyses.

Differences in demographic variables and microstructure behaviors by sex were examined using t-tests for continuous variables (e.g., fat mass index, bite rate) or χ^2for categorical variables (e.g., maternal education level).

2.3.3. Consistency of Microstructure Behaviors.

Consistency of children’s microstructure behaviors across the four meals was assessed using two-way mixed effects ICC with absolute agreement. Only participants with coded behavior across all four meals were included in ICC analyses (see Table 3 for sample sizes by behavior). Three children were excluded just from eating rate measures due to missing measured intake values.

2.3.4. Consistency of Microstructure-Microstructure Associations.

Intra-individual associations between microstructure behaviors across meals were assessed using repeated measures correlations (Bakdash & Marusich, 2017). The strength of associations was interpreted according to effect size categories for the absolute value of r: small < 0.30, medium 0.30 – 0.50, and large > 0.50 (Cohen, 1988).

2.3.5. Association Between Meal Microstructure and Intake.

General linear regressions with standardized coefficients were used to test the effect of meal microstructure on gram and energy intake for each meal separately. Not all children had complete microstructure data for all portion size meals (see section 2.3.3 Consistency of Microstructure Behaviors), therefore, sample sizes varied slightly for each meal (n = 83 – 89; see Table S4 for sample and microstructure descriptive statistics by meal and sex). All models adjusted for pre-meal fullness, sex, age, fat mass index, average liking of meal foods, and meal order. Following the procedure used in Fogel et al., (2017a) meal microstructure variables were entered in a single block to assess their relative effect on intake (i.e., all microstructure variables were included together in the model). Models included number of bites, number of sips, meal duration, bite size, eating rate, and percent active eating time with bite size and eating rate calculated using grams or kcal to match the intake measure being modeled. While meal microstructure variables can be correlated, there was no evidence of collinearity in any of the intake regression models for grams or kcal based on condition indices derived from eigenvalues (all condition indices were < 7). Multicollinearity is detected when condition indices > 10 (J. H. Kim, 2019) or >15 (Shrestha, 2020). While Fogel and colleagues examined associations with intake separately for faster and slower eaters, both groups showed a similar pattern of associations. Since eating rate has been associated with intake (Fogel et al., 2017a, 2017b), in the present study, we included it as a continuous variable to the model, thus extending work from Fogel et al., (2017a). Latency to first bite was not included in models due to limited variability (range: 0 – 1.9 min), possibly because all children were fasted for 3 hours prior to the meal and, therefore, had high motivation to eat. Other eating behaviors were not included if they were components of calculated scores to avoid collinearity (e.g., total eating time and active eating time both contribute to percent active eating time and, thus, were omitted). Relative importance analysis (Tonidandel & LeBreton, 2011) was also used to directly assess the relative effect of each eating behavior on intake.

2.3.6. Association Between Meal Microstructure and Adiposity.

Following analyses in Fogel and colleagues (2017a), general linear regressions with standardized coefficients were used to test the associations between meal microstructure behaviors and adiposity. Adiposity was indexed using both BMI percentile and fat mass index (fat mass, kg/height, m²). Separate models were conducted for each of the 8 eating behaviors, and each model included identical covariates (i.e., pre-meal fullness, sex, age, and average meal food liking). Within each portion size condition, model estimates for the 8 eating behaviors were adjusted using the Benjamini and Hochberg approach to constrain false discovery rate (FDR) < 0.05.

3. Results

3.1. Demographic Characteristics.

There were no sex differences in age, BMI percentile, race, income, or maternal education (ps > 0.225). Girls did, however, have higher fat mass index compared to boys (t(83) = −2.94, p = 0.004; Table 1). There were few sex differences in microstructure behaviors across portion size meals (see Table S4 for descriptive statistics by sex). Compared to boys, girls took more bites per minute of active eating time (i.e., bite rate, active) across all four meals (ps < 0.019) and more bites per minute of meal duration (i.e., bite rate) for portion sizes 2-4 (ps < 0.027). For portion size 2 only, girls took smaller bites (g and kcal per bite; than boys (ps < 0.037).

3.2. Consistency of Meal Microstructure.

Means and standard deviations for eating behaviors across portion size conditions are in in Table 3. Children showed moderate consistency across meals for the majority of eating behaviors. All coded behaviors except latency to first bite were moderately consistent across the four meals (ICCs > 0.53; Table 3). For the computed behaviors, bite rate had the greatest within-child consistency regardless of whether it was computed across meal duration or active eating time. Similarly, eating rate showed moderate within-child consistency across meals regardless of amount consumed (g or kcal) or duration (active eating time or meal duration). In contrast, bite size (g or kcal) and percent active eating time were not consistent across meals (ICCs < 0.33; Table 3). Together, this indicates that aside from bite size and active eating time, the majority of meal microstructure behaviors were consistent across meals that varied in portion size.

3.3. Consistency of Microstructure-Microstructure Associations

Repeated measures correlation was used to assess within-child (i.e., intra-individual) associations between eating behaviors across the four meals (Table 4; between-child correlations for all behaviors are reported for each meal in supplemental tables S5-S6). Overall, within-child associations had larger effect sizes for behaviors that were calculated using similar coded values (e.g., sips and sip rate). Similar patterns of association were also seen for behaviors computed using gram and energy intake. Across the four meals, there was a large effect for the associations between bite number, meal duration, and active eating time. While larger bite size and faster bite rate were associated with faster eating rates, the effects were stronger for bite size (g and kcal) compared to bite rate, regardless of whether this was calculated using meal duration or active eating durations. In contrast, there were small or very small effects for associations between sips and latency to first bite. Together, this suggests that there are consistent within-child patterns of eating behaviors across meals varying in portion size.

Table 4.

Repeated Measures (Within-Child) Correlations between Meal Microstructure

		1	2	3	4	5	6	7	8	9	10	11	12	13	14	15
Coded Behaviors	1. Bites, N
	2. Sips, N	.09
	3. Latency to 1st Bite, min	−.12	.04
	4. Meal Duration, min	.54	.21	−.01
	5. Active Eating Time, min	.66	.17	−.04	.81
Computed Behaviors	6. Bite Rate, bites/min	.50	−.10	−.17	−.34^	−.08
	7. Active Bite Rate, bites/min (active)	.41^	−.05	−.07	−.28	−.29	.85
	8. Sip Rate, sips/min	−.10	.83	.01	−.13	−.08	.00	−.01
	9. Active Sip Rate, sips/min (active)	−.13	.84	.07	−.06	−.14	−.10	.01	.95
	10. Bite Size, g/bite	−.40^	−.06	.06	−.23	−.24	−.34^	−.34^	−.004	.01
	11. Bite Size, kcal/bite	−.35^	−.20	.07	−.24	−.23	−.27	−.28	−.14	−.13	.92
	12. Eating Rate, g/min	−.24	−.09	−.10	−.63	−.40^	.37^	.23	.15	.06	.51	.48
	13. Eating Rate, kcal/min	−.14	−.25	−.07	−.55	−.34^	.43^	.31^	−.06	−.13	.30^	.52	.83
	14. Eating Rate, g/min (active)	−.35^	−.09	.02	−.53	−.55	.11	.22	.09	.10	.72	.67	.86	.69
	15. Eating Rate, kcal/min (active)	−.26	−.24	.03	−.51	−.50	.23	.30^	−.10	−.09	.55	.79	.74	.89	.84
	16. Percent Active Eating, active/meal duration	.26	−.04	−.14	−.22	−.36^	.49^	−.002	.10	−.13	−.14	−.11	.34^	.33^	−.14	−.07

g: grams, ICC: intraclass correlation coefficient; kcal: kilocalories

Effect size categories for absolute value of r: small < 0.30; medium 0.3-0.5; large > 0.50. Large effects are bolded and medium effects are denoted with ^.

3.4. Association Between Meal Microstructure and Intake

After adjusting for covariates (see 2.3.5), greater number of bites, longer meal duration, and faster eating rates (g and kcal) were each independently associated with greater intake across all portion size conditions (Table 5). This suggests that each of these behaviors contributed independently to total meal intake (g and kcal). Relative importance analyses showed that the effects of number of bites, meal duration, and eating rate were similar on total energy and weight of food consumed, regardless of portion size condition. On the other hand, bite size was not consistently associated with meal intake in grams or kcal. The effect of bite size on intake was 1.5-10 times smaller compared to the independent effects from total bites, meal duration, and eating rate (Table 5). Together, this indicates that more bites, longer meal duration, and faster eating rate characterize a style of eating that contributes to greater intake across meals that vary in portion size.

Table 5.

Effect of Microstructure on Intake

Intake, g
	Portion 1		Portion 2		Portion 3		Portion Size 4
	β (SE)	RW	β (SE)	RW	β (SE)	RW	β (SE)	RW
Pre-Meal Fullness	−0.006 (0.04)	0.69	−0.01 (0.06)	3.54	−0.03 (0.04)	3.89	−0.03 (0.05)	1.31
Sex (ref=male)	−0.17 (0.09)	1.36	−0.11 (0.12)	2.60	−0.02 (0.08)	0.42	−0.18 (0.10)	0.92
Age, years	−0.01 (0.04)	0.33	0.01 (0.06)	1.06	−0.02 (0.04)	0.78	0.01 (0.05)	0.77
Fat Mass Index	0.06 (0.04)	1.70	0.04 (0.06)	1.04	0.10 (0.04)**	3.27	0.08 (0.05)	1.08
Average Liking	0.04 (0.04)	1.93	0.04 (0.06)	1.63	0.05 (0.04)	1.13	0.05 (0.05)	1.39
Bites, N	0.50 (0.09)***	29.75	0.21 (0.11)+	22.01	0.40 (0.10)***	28.42	0.17 (0.07)*	19.40
Sips, N	0.07 (0.05)	10.39	0.08 (0.06)	8.55	0.004 (0.04)	6.54	0.15 (0.05)**	14.31
Meal Duration, min	0.58 (0.09)***	23.93	0.84 (0.12)***	29.44	0.70 (0.09)***	27.67	0.83 (0.08)***	29.80
Bite Size, g/bite	0.26 (0.07)**	8.26	0.09 (0.10)	6.96	0.21 (0.08)**	7.19	<0.001 (0.07)	2.89
Eating Rate, g/min	0.55 (0.08)***	17.10	0.74 (0.10)***	20.91	0.57 (0.08)***	17.05	0.80 (0.08)***	24.46
% Active Eating, active/meal duration	0.06 (0.05)	4.57	−0.02 (0.06)	2.27	0.08 (0.04)+	3.63	0.16 (0.07)*	3.68
Intake, kcal
	Portion 1		Portion 2		Portion 3		Portion Size 4
	β (SE)	RW	β (SE)	RW	β (SE)	RW	β (SE)	RW
Pre-Meal Fullness	0.09 (0.06)	0.68	0.02 (0.06)	4.98	0.01 (0.04)	4.49	−0.02 (0.05)	1.26
Sex (ref=male)	−0.17 (0.09)	2.03	−0.11 (0.12)	1.25	−0.11 (0.08)	0.82	−0.12 (0.11)	1.03
Age, years	−0.04 (0.04)	1.08	0.01 (0.06)	0.90	−0.03 (0.04)	1.08	−0.05 (0.05)	1.55
Fat Mass Index	0.02 (0.04)	0.78	0.10 (0.06)+	3.12	0.14 (0.04)**	4.57	0.07 (0.06)	1.54
Average Liking	0.06 (0.04)	2.26	0.01 (0.06)	1.01	0.06 (0.04)	1.07	0.05 (0.05)	1.72
Bites, N	0.56 (0.09)***	33.79	0.38 (0.12)**	29.23	0.44 (0.09)***	27.89	0.21 (0.08)*	23.00
Sips, N	−0.03 (0.05)	3.08	−0.01 (0.06)	1.01	−0.03 (0.04)	1.36	0.04 (0.05)	2.11
Meal Duration, min	0.55 (0.09)***	20.27	0.64 (0.13)***	22.14	0.54 (0.08)***	21.16	0.81 (0.09)***	32.76
Bite Size, kcal/bite	0.27 (0.08)	9.20	0.18 (0.10)+	9.18	0.23 (0.08)**	10.78	0.01 (0.07)	4.17
Eating Rate, kcal/min	0.60 (0.08)***	20.77	0.69 (0.11)***	25.39	0.59 (0.09)***	21.26	0.88 (0.08)***	32.76
% Active Eating, active/meal duration	0.06 (0.05)	6.07	−0.01 (0.07)	1.67	0.07 (0.05)	5.51	0.13 (0.05)*	3.60

β: standardized coefficient, g: grams, kcal: kilocalories, ref: reference category, RW: standardized relative weight Portion 1: 769 g, 1048 kcal; Portion 2: 1011.4 g, 1376.8 kcal; Portion 3: 254.8 g, 1706.66 kcal; Portion 4: 1499.2 g, 2037.4 kcal

⁺p < 0.100,

^*p < 0.050,

^**p < 0.010,

^***p < 0.001

Total weight and energy consumed increased with increasing portion size (reported in (Keller et al., 2023), therefore, it is unsurprising to see that the behaviors that contributed to intake within a meal (e.g., number of bites, meal duration, and eating rate) also increased with amount of food served (see supplemental materials and Table S9). The high consistency of these behaviors across meals (see 3.2 Consistency of Meal Microstructure) and their consistent association with intake within each meal suggest that while portion size influences these behaviors, the children who took the most bites in portion size 1 were also likely to take the most bites in portion size 4.

3.4.1. Sensitivity Analyses.

Sensitivity analyses were conducted to determine if the time of day (lunch versus dinner) or the modality of book reading (by research assistant pre-COVID versus audiobook post-COVID) impacted the effect of meal microstructure on intake. Both meal time and book modality were added to the models as covariates and did not impact the pattern or significance of results for g or kcal across the for portion size meals (see supplemental materials and Table S10). Separately, the effect of plate cleaning was examined by removing the two participants who consumed ≥ 95% of the amount served in portion size 1. Removing these children did not impact the pattern or significance of results (see supplemental materials and Table S11). No children consumed ≥ 95% of the amount served for portion sizes 2-4.

3.5. Association Between Meal Microstructure and Adiposity

To understand the relationship between meal microstructure and adiposity, associations with BMI percentile and fat mass index were tested. After adjusting for pre-meal fullness, sex, age, fat mass index, average liking of meal foods and meal order, eating rate was positively associated with BMI percentile across all meals and these results survived FDR correction (Figure 2A, Table 6). In contrast to the consistent association between eating rate and weight status, bite size was not consistently associated with weight status across portion conditions after correction for FDR (Table 6). Greater fat mass index was only associated with lower percent of active eating time during the largest portion size meal (Figure 2B, Table 6). To summarize, BMI percentile showed stronger associations with eating behaviors than measured fat mass index.

Figure 2.

Effect of child adiposity on eating behaviors. A) Standardized coefficients for the effect of BMI percentile across portion size meals; B) Standardized coefficients for the effect of fat mass index across portion size meals. Color gradient: darkest for portion size 1 meal → lightest for portion size 4 meal. Bolded bars: unadjusted p<0.05. FDR corrected p-values: +p<0.01, *p<0.05. **p<0.01. Portion 1: 769 g, 1048 kcal; Portion 2: 1011.4 g, 1376.8 kcal; Portion 3: 254.8 g, 1706.66 kcal; Portion 4: 1499.2 g, 2037.4 kcal

Table 6.

Association between Adiposity and Meal Microstructure Adjusted for Age, Sex, Average Food Liking and Pre-Meal Fullness

BMI Percentile
	Portion 1		Portion 2		Portion 3		Portion 4
	β (SE)	PFDR	β (SE)	PFDR	β (SE)	PFDR	β (SE)	PFDR
Bites, N	−0.06 (0.11)	0.659	−0.06 (0.11)	0.538	0.01 (0.10)	0.919	−0.07 (0.11)	0.636
Sips, N	0.09 (0.11)	0.542	−0.13 (0.11)	0.254	0.11 (0.11)	0.401	−0.07 (0.11)	0.636
Meal Duration, min	−0.20 (0.10)+	0.097+	−0.19 (0.10)+	0.116	−0.14 (0.10)	0.296	−0.18 (0.11)	0.236
Bite Size, g/bite	0.21 (0.11)*	0.094+	0.14 (0.11)	0.214	0.18 (0.10)+	0.238	0.02 (0.11)	0.877
Bite Size, kcal/bite	0.22 (0.11)*	0.094+	0.29 (0.11)**	0.016*	0.16 (0.11)	0.279	0.07 (0.11)	0.636
Eating Rate, g/min	0.39 (0.10)***	0.001**	0.33 (0.10)**	0.005**	0.27 (0.10)*	0.065+	0.23 (0.11)*	0.137
Eating Rate, kcal/min	0.40 (0.10)***	0.001**	0.41 (0.10)***	0.001**	0.25 (0.10)*	0.065+	0.33 (0.11)**	0.024*
Percent Active Eating, active/meal duration	0.02 (0.11)	0.867	−0.18 (0.11)	0.174	−0.04 (0.11)	0.855	−0.17 (0.11)	0.236
Fat Mass Index
	Portion 1		Portion 2		Portion 3		Portion 4
	β (SE)	PFDR	β (SE)	PFDR	β (SE)	PFDR	β (SE)	PFDR
Bites, N	−0.04 (0.11)	0.834	−0.07 (0.11)	0.58	−0.10 (0.11)	0.996	−0.10 (0.11)	0.697
Sips, N	0.15 (0.11)	0.357	−0.06 (0.12)	0.58	0.11 (0.11)	0.488	−0.02 (0.12)	0.845
Meal Duration, min	−0.06 (0.11)	0.796	−0.12 (0.11)	0.367	−0.001 (0.11)	0.996	−0.02 (0.11)	0.845
Bite Size, g/bite	0.17 (0.11)	0.351	0.21 (0.11)+	0.108	0.15 (0.11)	0.415	0.05 (0.12)	0.845
Bite Size, kcal/bite	0.13 (0.11)	0.384	0.24 (0.11)*	0.078+	0.12 (0.11)	0.488	0.05 (0.12)	0.845
Eating Rate, g/min	0.24 (0.11)*	0.18	0.27 (0.11)*	0.077+	0.17 (0.11)	0.415	0.012 (0.11)	0.697
Eating Rate, kcal/min	0.23 (0.11)*	0.18	0.26 (0.11)*	0.077+	0.19 (0.11)+	0.415	0.15 (0.12)	0.697
Percent Active Eating, active/meal duration	0.002 (011)	0.985	0.18 (0.11)	0.175	−0.10 (0.11)	0.488	−0.35 (0.11)**	0.010*

β: standardized coefficient, FDR: false discovery rate corrected p-value; g: grams, kcal: kilocalories,

Portion 1: 769 g, 1048 kcal; Portion 2: 1011.4 g, 1376.8 kcal; Portion 3: 254.8 g, 1706.66 kcal; Portion 4: 1499.2 g, 2037.4 kcal

⁺p < 0.100,

^*p < 0.050,

^**p < 0.010,

^***p < 0.001

4. Discussion

Overall, this study provides evidence that eating behaviors are generally consistent and that there is an ‘obesogenic’ style of eating that is consistently associated with greater intake across meals. In line with prior observations, all measured eating behaviors except bite size and percent active eating were consistent within children across meals, despite variations in the amount served, suggesting that these eating behaviors may be stable behavioral traits. In addition to within child stability, the combination of faster eating rates, greater bite number, and longer meal duration consistently predicted intake across the meals. This partially replicates work from the GUSTO cohort (Fogel et al., 2017a) in Singapore by demonstrating that an obesogenic style of eating can also be observed in children from the United States. Each of these eating behaviors was independently associated with greater food consumption (kcal and g). Additionally, higher BMI percentile was associated with faster eating rates and larger bite sizes. Together, this provides evidence for the consistent impact of eating behaviors across meals varying in portion size and suggests that ‘obesogenic’ styles of eating are stable behavioral traits.

With the exception of latency to first bite, bite size, and percent active eating time, children’s eating behaviors were generally consistent across meals. In addition, within-child (i.e., intra-individual) associations between eating behaviors across the four meals showed similar patterns of associations between behaviors as previously reported in the literature. As seen in the GUSTO cohort, the number of bites taken was negatively associated with bite size while faster eating rate was positively associated with bite size and bite rate (Fogel et al., 2017a). These patterns of association did not differ when behaviors were computed using food energy or weight, however, this should be further examined in meals that differ in overall energy density and composition. Together, this suggests that there are consistent within-child patterns of eating behaviors across meals varying in portion size.

Child eating behaviors were also consistently associated with intake across meals. Greater number of bites, longer meal duration, and faster eating rates were independently associated with greater intake across all portion size conditions. This replicates findings from the GUSTO cohort showing faster eating rate was associated with greater energy consumption at a single meal (Fogel et al., 2017a, 2017b). While larger bite size was also related to greater consumption in the GUSTO cohort (Fogel et al., 2017a), we found that this association was less consistent in the present study. A key distinction between the two studies was our inclusion of demographic covariates in the models, which may have reduced power to find a significant effect of bite size on intake in all portion size meals. In particular, the present study adjusted for sex and adiposity, both of which impact energy needs. While a systematic review showed no sex differences in sensorimotor control of mastication prior to ages 10-12-years-old (Almotairy et al., 2018), there is evidence for sex differences in eating behaviors both in the current study and the study by Fogel and colleagues (2017). Therefore, adjusting for sex could have accounted for more variability in bite size in the present study. Additionally, Fogel and colleagues (2017) ran independent models for faster and slower eaters while the present study included all children in the same model and added eating rate as an additional predictor of intake. As bites size was associated with eating rate, it is possible that including eating rate in the model accounts for some of the effect of bite size on intake. Despite these differences across studies, this pattern of results strengthens the evidence base for an ‘obesogenic’ style of eating that contributes to greater food consumption across meals.

This ‘obesogenic’ style of eating, and in particular, faster eating rates, may contribute to greater food intake due to reduced oro-sensory (i.e., ‘in-mouth’) exposure to food, which may delay satiation. Oro-sensory exposure is determined by both duration of bites (i.e., total time in mouth) and intensity of oral processing (e.g., chewing) (Lasschuijt et al., 2021). Within the GUSTO cohort, children who ate faster took fewer chews and had less oral processing time per bite compared to slower eaters (Fogel et al., 2017a). Additionally, greater amount of chewing and longer total oral processing time were associated with lower energy intake in the GUSTO cohort (Fogel et al., 2017a). Greater mastication may lead to greater satiety by influencing post-prandial metabolic or hormonal responses (e.g., GLP-1, ghrelin, CCK) (Hollis, 2018). While the present study was not able to directly code chews, faster eating rates suggest that children had reduced oral processing of food. Reduced oral processing could delay the development of satiation (Lasschuijt et al., 2021) which may also lead to longer meal duration. Indeed, longer meal duration and faster eating rate were each associated with greater intake after adjusting for the impact of the other variable (i.e., meal duration was associated with intake after adjusting for eating rate). Future studies are needed to determine if these two behaviors – eating rate and meal duration – have additive effects or independent effects on children’s intake. Together, this suggests that reduced oro-sensory processing may be one mechanism through which this style of eating contributes to greater meal intake.

In addition to contributing to greater intake, the ‘obesogenic’ style of eating has also been associated with child adiposity and weight status (Fogel et al., 2017b; Llewellyn et al., 2008; Pearce et al., 2022). Similar to prior work, we showed that children with higher BMI percentile tended to eat faster and consumed more food per minute. We also replicated the association between weight status and bite size seen in the GUSTO cohort, but these effects were only observed at the smaller portion size conditions. In contrast, there was no relationship between bite size and weight status in the larger portion size conditions, possibly because these meals deviated too far from a typical child meal (i.e., the reference portion size condition). Presenting double the expected amount of food may cause children who typically take smaller bites to increase their bite size (Fisher, 2007; Fisher et al., 2003; Gómez-Zúñiga & Wintergerst, 2023), thus, reducing the association with weight status when larger portions are served. Associations between weight status and bite size have received less attention compared to bite number and eating rate (Pearce et al., 2022) and findings in the literature are mixed. Within the GUSTO cohort, children with overweight or obesity took larger bites relative to those with healthy weight during a self-served meal (Fogel et al., 2017a), however, another study found no association between bite size of a snack and weight status among 8-10-year-olds (Gómez-Zúñiga & Wintergerst, 2023). While the overall pattern of results was consistent for fat mass index, the associations were generally weaker for this DXA-derived measure compared to BMI. Similarly, although children with overweight or obesity in the GUSTO cohort ate faster than those with healthy weight, findings related to adiposity were mixed (Fogel et al., 2017b). Using MRI, Fogel and colleagues (2017b) showed that faster eaters had greater subcutaneous, but not visceral, adipose tissue compared to slower eaters. Together, this suggests a need for future work to better understand the association between ‘obesogenic’ patterns of eating behaviors and body composition.

This study was novel in its ability to characterize eating behaviors across meals that varied in portion size. However, these findings must be considered in the context of a few limitations. First, while we were able to establish consistency of child eating behaviors across variations in portion size, the foods served were the same for all meals. Since eating behaviors were measured in meals that only varied in portions served, it remains unclear how textural differences in foods, eating utensils, or other contextual differences may impact the consistency of children’s oral processing behaviors. While the meals in this study included foods that varied in texture (e.g., macaroni and cheese versus grapes), all meals contained identical foods. Therefore, additional studies are needed to assess the impact of meal context and food texture on consistency of meal eating behaviors in children. Additionally, all children in the present study were without obesity so it remains unclear if associations seen with BMI percentile or adiposity will be replicated in children with obesity. Lastly, this sample had limited racial and ethnic diversity. While the sample reflected the community in Central, PA, it is unclear whether findings would generalize to more urban environments or to other ethnic or racial groups. Despite these limitations, this study provided initial evidence that children’s eating behaviors are consistent predictors of intake across meals varying in portion size. Given this stability, eating behaviors may be a promising target for interventions aiming to reduce overconsumption and obesity.

In conclusion, children’s eating behaviors may represent a consistent behavioral trait that can be targeted to reduce food consumption at meals. Faster eating rates, longer meal duration, and greater number of bites were independently associated with greater intake (gram and kcal) across meals, replicating their contribution to an ‘obesogenic’ style of eating. Similarly, we replicated previous associations (Fogel et al., 2017b) showing positive associations between eating rate and weight status (i.e., BMI percentile). The stability of meal microstructure highlights the potential of targeting these behaviors with interventions to reduce overconsumption and obesity. While there have been inconsistent effect sizes for interventions that have only targeted eating speed (Cox et al., 2022), the current findings suggest that interventions focused on altering an obesogenic style of eating may be lead to more impactful and sustained results. Together, this highlights the potential of meal microstructure as a target in the prevention of excess consumption and pediatric obesity.