Association of Snacking Frequency, Size, and Energy Density with Weight Status among Preschool-Aged Children in the United States

Abstract

Background

Snacking (ie, eating between meals) is common among US preschool-aged children, but associations with weight status are unclear.

Objective

This research evaluated associations of snack frequency, size, and energy density as well as the percent of daily energy from snacking with weight status and sociodemographic characteristics among US children aged 2 to 5 years.

Design

Cross-sectional analysis of 2007–2018 National Health and Nutrition Examination Survey data using two, caregiver proxy, 24-hour dietary recalls.

Participants/setting

US children aged 2 to 5 years (n = 3,313) with at least one snack occasion over 2 days of intake.

Main outcome measures

Snacking parameters included frequency (number of occasions per day), size (kilocalories per occasion), and energy density (kilocalories per gram per occasion) as well as percent of daily energy from snacking.

Statistical analyses

Generalized linear regression models evaluated associations of snacking with child weight status (ie, normal weight and overweight/obesity), adjusting for survey weights, energy misreporting, mean meal size, and sociodemographic covariates.

Results

Children with overweight/obesity consumed more frequent snacks (2.8 [0.06] vs 2.5 [0.03] snacks/day, respectively; P < 0.001), larger snacks (188 [4] vs 162 [23] kcal/occasion, respectively; P < 0.001), and a greater percent of daily energy from snacking (29.80% [1.00%] vs 26.09% [0.40%], respectively; P < 0.001) than children with normal weight. Mean snack frequency and size as well as percentage of daily energy from snacking varied with child age, gender, and head of household education. Associations of snacking with child race and ethnicity were less consistent.

Conclusions

These nationally representative findings provide evidence that the consumption of larger, more frequent snacks is associated with overweight/obesity among US children aged 2 to 5 years and snacking varies by sociodemographic characteristics.

Snacking (ie, eating between meals) is nearly universal among young children.^1,2 Although snacks are often considered an accessory part of the diet, snacking contributes significant energy to young children’s diets.³ During 2017–2018, snacking provided 28% to 29% of daily energy among US children aged 2 to 5 years—more energy than at any other single meal.^4–6 Snacking among young children is not only associated with higher intakes of daily energy intake,⁷ but also higher intakes of added sugars,⁸ accounting for close to 40% of total daily intake of added sugars among US children aged 2 to 5 years.⁹ Current levels of snacking among US children represent increases from previous decades in terms of the number of children who snack daily, the number of snacks consumed daily, and the contributions of snacking to daily energy intake.^10,11 The largest increases in per capita snacking intake among children have been observed in non-Hispanic Black children and in groups with the lowest levels of family income and education.¹⁰ Collectively, these observations have generated concerns about the role of snacking in promoting excessive intake and increasing obesity (OB) risk among children, particularly among populations for whom lower diet quality,^12–15 higher burdens of obesity,^16–18 and targeted marketing of unhealthy snack food and beverages has been observed.¹⁹

Contributions of snacking to weight status among young children are not well characterized. As reviewed by Larson and Story,²⁰ only a handful of studies to date have evaluated associations of snacking with weight status among preschool-aged children and have produced equivocal findings. For example, a greater snacking frequency was associated with higher waist circumference among preschool-aged children from India²¹; higher snack energy intake was associated with lower odds of overweight (OW) among preschool-aged children from France²²; and the association of snacking with weight status was not statistically significant among preschool-aged children from Japan.^23,24 In the only US population-representative study to date, children aged 1 to 5 years with OW/OB tended to snack more frequently than children with normal weight (NW).²⁵ Subsequent research has shown that the larger snack sizes (eg, more calories per snack) and higher energy-density of snacks consumed by young children are associated with lower diet quality among US preschool-aged children.²⁶ To our knowledge, snacking size and energy-density associations with weight status among young children have not been systematically evaluated in a nationally representative sample.

The primary objective of this research was to address these gaps by evaluating associations of snack frequency, size, and energy density as well as the percent of daily energy from snacking with weight status among US preschool-aged children. Based on associations observed among older children,²⁷ we hypothesized that greater snack frequency and larger snack sizes would be associated with higher weight status among US preschool-aged children. A secondary objective was to evaluate sociodemographic correlates of snacking.

METHODS

Design

A cross-sectional analysis of snacking among US preschool-aged children was conducted using 2007–2018 National Health and Nutrition Examination Surveys (NHANES) data. NHANES is an ongoing, nationally representative study of the nutritional and health status of the civilian, noninstitutionalized US population.²⁸ NHANES uses a complex, multistage, probability sampling design with county as the primary sampling unit from which clusters of households and participants are randomly selected.^29,30 NHANES collects informed consent along with detailed data on nutritional status via dietary assessments, anthropometric measurements, laboratory tests, and clinical examinations. Child snacking was assessed using two 24-hour dietary recalls and child weight status was assessed using objectively measured heights and weights. For young children aged 5 years or younger, parents or a proxy familiar with the child’s intake provided the dietary recalls. The first 24-hour recall was collected in person at the NHANES Mobile Examination Center. A second 24-hour recall was collected via telephone 3 to 10 days after the initial 24-hour recall.²⁹ Survey weights adjusting for dietary-interview-specific nonresponse and the day of the week were applied. The National Center for Health Statistics Research Ethics Review Board approved the study protocols. Other design and data collection procedures for NHANES are detailed elsewhere.^28,31 This project was determined to be exempt from review by the Temple University Institutional Review Board.

Participants

Participants were children aged 2 to 5 years of age. Exclusion criteria were fewer than 2 days of dietary intake (n = 1,859), missing height or weight data needed to calculate body mass index (n = 109), a diabetes diagnosis (n = 4), use of medications known to affect weight status (n = 47), consumption of breast milk due to lack of nutritional data (n = 27), and missing head of household (HH) education data used as a covariate (n = 97). Given the focus of this analysis on snacking frequency, size, and energy density, 52 cases were removed without any snacking occasions (ie, nonsnackers; 1.5% of the sample), leaving a final sample of 3,313 children for analyses of snacking and weight status.

Measures

Data from eight NHANES dataset files were used in this research: demographic files, body measures files (ie, anthropometrics), Day 1 and Day 2 dietary interview individual foods files; Day 1 and Day 2 total nutrient intakes files; and diabetes and prescription drug files (used for exclusion criteria).^32–37 The measures included in this study are described below.

Dietary intake.

Children’s dietary intake was collected by adult-proxy using the US Department of Agriculture What We Eat in America Survey’s 5-step Automated Multiple-Pass Method.³⁰ The mean of the 2 days was used to construct all dietary variables of interest. Eating occasions were participant defined using a pre-determined list (eg, “breakfast”). In this analysis, snacking occasions were those identifed by participants as “snacks,” “beverages” not otherwise included in meals, or “extended consumption,” as well as the Spanish equivalents of “merienda,” “bebida,” “bocadillo,” “tentempie,” and “entre comida” to capture eating in between meals. Foods/beverages consumed at the same time of day were aggregated. Consistent with previous work done by this team, snacking occasions providing trivial energy (ie, <5 kilocalories) were excluded.^25,27 Four snacking parameters were calculated: mean snack frequency (number of daily occasions), mean snack size (kilocalories per occasion), mean snack energy-density (kilocalories per gram per occasion), and the percent of daily of energy from snacking. Meal occasions were those identifed by participants as “breakfast,” “lunch,” “dinner,” “supper,” “brunch,” “other,” and the Spanish equivalents “desayuno,” “almuerzo,” “comida,” and “cena” from which mean meal size was calculated. Mean meal size (kilocalories per occasion) was included as a covariate in the main analyses to take into account how much children eat at meals when evaluating associations of snacking with weight status.

Following the methods of Murakami and Livingstone³⁸ and consistent with our previous research,²⁷ the ratio of energy intake to estimated energy expenditure (EI:EER) was used to adjust models for dietary reporting accuracy without biasing sample selection.^39,40 EER was estimated using dietary reference intake equations based on age, gender, height, weight, and physical activity level.⁴¹ An assumption of “low active” level of physical activity (≥1.4 to <1.6) was made for all children based on NHANES accelerometer data for children aged 6 years and older.⁴² This approach is consistent with previous work³⁸ and takes into account the limited data on physical activity among preschool-aged children, which reveals high rates of sedentary behavior among children in center-based childcare.⁴³

Weight Status.

Height was measured in centimeters using a stadiometer and weight was measured in kilograms using a digital scale during an in-person mobile examination center exam.⁴⁴ Centers for Disease Control and Prevention growth charts⁴⁵ were used to derive age-and-gender specifc body mass index percentiles using the “agd” package in R software.⁴⁶ Weight status was classified using standard criteria: NW <85th percentile, OW ≥85th to <95th percentile, and OB ≥95th percentile; OW and OB categories were combined in analyses.⁴⁷

Sociodemographic characteristics.

Child age (years), gender (male or female), race, ethnicity, HH age (<40 or ≥40 years), education (less than high school, high school/some college, or college graduate or above), marital status (partnered vs not partnered), and poverty-to-income ratio (PIR) were evaluated in the analyses. Sociodemographic data were collected by self-report, via adult-proxy using the NHANES computer-assisted personal interviewing system.²⁹ Race and ethnicity categories included Hispanic (includes Mexican-American and other Hispanic), non-Hispanic Black, non-Hispanic White, and Other (includes non-Hispanic Asian and other race, including multiracial). Non-Hispanic Asian is included in the Other category because this analysis includes data from two survey cycles before 2011 when NHANES began to oversample Asian Americans to provide separate estimates for this group.³¹ Family PIR was calculated as a percentage of the poverty threshold and classified as a dichotomous variable as <125% and ≥125% of the poverty guidelines.⁴⁸

Statistical Analyses

Statistical analyses were performed using R version 3.6.1,⁴⁹ including applicable survey weights to account for the NHANES complex survey design.⁵⁰ Sample weights were created using NHANES analytical guidelines (the original 2-year dietary 2-day sample weight divided by six) for the six cycles (12 years) of NHANES data combined in this analysis.⁵¹ Descriptive statistics were generated for all variables of interest; sociodemographic characteristics were included for children with and without snacking occasions and expressed as counts and percentages for categorical variables and means and standard errors (SE) for continuous variables.

Generalized linear regression models were used to evaluate associations of snacking parameters (ie, frequency, size, energy density, and percent of daily energy) with weight status (primary objective). Separate regression models were used for each of the four snacking parameters with adjustments for EI:EER ratio, mean meal size, and sociodemographic covariates. Each of the three separate models evaluating snacking frequency, size, and energy density respectively were adjusted for the two other snacking dimensions; for instance, the model evaluating snack frequency was adjusted for snack size and energy density. Consistent with previous studies of dietary energy density, separate models were used to evaluate mean snack energy density based on snack foods and beverages, snack foods only, and beverages only.⁵² Predictive marginal means, or probability-weighted averages, are presented, with SE for each snacking parameter by weight group to allow comparisons between group outcomes while controlling for different covariate distributions in the groups.⁵³ Beta coefficients are presented, with SE and 95% CI, to estimate unit change in the outcome variable due to a 1-unit increase in the corresponding independent variable.

Consistent with previous analyses of snacking among US children,^10,27 child race, ethnicity, gender, and age as well as HH age, education, marital status, and family PIR were considered as potential covariates in models predicting weight status (secondary objective). Given relatively limited research to date on the specific snacking parameters of interest (e.g., snack size), a data-driven, backwards stepwise elimination approach (using lowest Akaike Information Criterion (AIC) value as the inclusion threshold for model selection⁵⁴) was used to evaluate and retain the following covariates: child race and ethnicity, gender, age, and HH education. For all tests, standard errors were estimated by Taylor Series Linearization^55,56 and P values < 0.05 were used to infer statistical significance.

RESULTS

Participant Characteristics

Participant characteristics are described in Table 1. Of the 3,313 children in the analytic sample, the mean age was 3.5 (0.03) years, with 53% of children identified as non-Hispanic White, 13% non-Hispanic Black, 24% Hispanic, and 10% identified as Other. Most children, 77%, had NW and 23% had OW/OB. Of caregivers, most were younger than age 40 years (72%), and 18% of HH reported less than a high school education, 52% reported high school or some college education, and 31% reported graduating college or above. A majority of HH were partnered (82%) (n = 3,224) and a little more than half (65%) of HH were at or above a PIR of 125% (n = 3,106).

Table 1.

Sociodemographic and anthropometric characteristics of 3,313 US children aged 2 to 5 years with at least one snacking occasion and 52 US children aged 2 to 5 years without at least one snacking occasion participating in National Health and Nutrition Examination Survey 2007–2018^a

Characteristic	Snackers	Nonsnackers
Child
Age (y)	3.48 (0.03)	3.77 (0.20)
BMIb	16.45 (0.04)	16.25 (0.37)
BMI z score	0.28 (0.03)	0.35 (0.22)
Gender
Female	1,625 (49)	27 (58)
Male	1,688 (51)	25 (42)
Race and ethnicity
Non-Hispanic White	1,077 (53)	9 (29)
Non-Hispanic Black	738 (13)	25 (34)
Hispanic	1,078 (24)	10 (27)
Otherc	420 (10)	8 (11)
Weight status d
Normal weight	2,582 (77)	45 (86)
Overweight/obese	731 (23)	7 (14)
Head of household
Age (y)
<40	2,421 (72)	42 (84)
≥40	892 (28)	10 (16)
Education
Less than high school	825 (18)	11 (16)
High school/some college	1,735 (52)	31 (72)
College graduate or above	753 (31)	7 (11)
Marital status e
Partnered	2,446 (82)	35 (74)
Not partnered	778 (18)	16 (26)
Poverty income ratio f g
<125%	1,460 (35)	25 (66)
≥125%	1,646 (65)	17 (34)

^aData were derived from demographic and body measures files, National Health and Nutrition Examination Survey 2007–2018, weighted.^32–37

^bBMI = body mass index.

^cIncludes non-Hispanic Asian and other race, including multiracial.

^dBMI-for-age percentiles: normal weight <85th, overweight ≥85th to <95th, and obese ≥95th percentile.⁴⁵

^eSnackers n = 3,224; nonsnackers n = 51.

^fSnackers n = 3,106; nonsnackers n = 42.

^gCalculated by dividing family income by the poverty guidelines specific to the survey year.⁴⁸

Associations of Snacking Frequency, Size, and Energy Density with Weight Status

Mean snack frequency was 2.6 (0.03) daily occasions, mean snack size was 168 (3) kcal/occasion, mean snack energy-density was 1.7 (0.02) kcal/g/occasion based on all foods and beverages, and percent of daily energy from snacking was 26.93% (0.33%) among US children aged 2 to 5 years. Table 2 gives predictive marginal means, beta coefficients, and 95% CIs for each snacking parameter by child weight status. Mean snack frequency and snack size as well as the percent of daily energy from snacking differed by child weight status. Children with OW/OB had, on average, 0.34 more snacking occasions daily compared to children with NW (2.8 [0.06] occasions/day vs 2.5 [0.03)] occasions/day, respectively; β = 0.34 [0.06]; P < 0.001). Children with OW/OB consumed, on average, 26 more calories per snacking occasion than children with NW (188 [4] kcal/occasion vs 162 [3] kcal/occasion, respectively; β = 26.35 [4.11]; P < 0.001). Children with OW/OB consumed, on average, 4% more daily energy from snacking (29.80% [1.00%] vs 26.09% [0.40%], respectively; P < 0.001) than children with NW. The association of snack energy density with weight status was not statistically significant, whether calculated based on foods and beverages (β = 0.03 [0.05]; P = 0.61) or foods alone (β = 0.00 [0.07]; P = 1.00). However, lower snack energy density was weakly associated with higher weight status when calculated with beverages only (β = −0.03 [0.01]; P < 0.05), such that mean energy density of beverage-only snacks was 0.03 kcal/g/occasion lower among children with OW/OB compared with children with NW children.

Table 2.

Mean snack frequency, snack size, snack energy density, and percent of daily energy from snacks associations with weight status among 3,313 US children aged 2 to 5 years participating in National Health and Nutrition Examination Survey 2007–2018 with at least one snacking occasion^a

	Weight statusb
Variable	All Children (N = 3,313)	Normal Weight (n = 2,582)	Overweight/obese (n = 731)
				β (SE)	95% CI
Snack frequency (occasion/person/d) c	2.57 (0.03)	2.49 (0.03)	2.84 (0.06)	0.34 (.06)	0.22 to 0.47***
Snack size (kcal/occasion) d	167.76 (2.77)	161.81 (2.53)	188.16 (4.42)	26.35 (4.11)	18.28 to 34.41***
Snack energy-density (kcal/g/occasion) e
From all foods and beverages	1.72 (0.02)	1.72 (0.02)	1.74 (0.05)	.03 (.05)	−.07 to .13
From foods alone (n = 3,270)	3.07 (0.03)	3.07 (0.03)	3.07 (0.07)	.00 (.07)	−.13 to .13
From beverages alone (n = 2,637)	0.48 (0.01)	0.49 (0.01)	0.46 (0.01)	−.03 (.01)	−.06 to −.01*
Percent of daily energy from snacks	26.93 (0.33)	26.09 (0.40)	29.80 (1.00)	3.71 (.66)	2.40 to 5.03***

^aData were derived from demographic characteristics, body measures, and What We Eat in America day 1 and 2 dietary intake files, National Health and Nutrition Examination Survey 2007–2018, weighted.^32–37

^bAll models were adjusted for child gender, child race and ethnicity, child age, ratio of energy intake to estimated energy requirement, head of household education, and mean meal size. BMI-for-age percentiles: normal weight <85th, overweight ≥85th to <95th, and obese ≥95th percentile.⁴⁵

^cModel adjusted for snack size and snack energy density.

^dModel adjusted for snack frequency and snack energy density.

^eModels adjusted for snack size and snack frequency.

^*P < 0.05.

^***P < 0.001.

Associations of Sociodemographic Characteristics with Snacking Frequency, Size, and Energy Density

As shown in Table 3, older children consumed more daily snacking occasions (β = .13 [0.02]; P < 0.001), more calories per snacking occasion (β = 22.44 [1.64]; P < 0.001), and a greater percent of energy from snacking (β = 1.64 [.20]; P < 0.001) compared with younger children. Similarly, male children consumed more daily snacking occasions (β = .34 [.05]; P < 0.001), more calories per snacking occasion (β = 29.53 (3.91); P < 0.001), and a greater percent of daily energy from snacking (β = 2.67 [.44]; P < 0.001) than female children. Non-Hispanic Black children consumed fewer daily snacking occasions (β = −.16 [0.05]; P < 0.01) but more calories per snacking occasion (β = 18.36 [4.80]; P < 0.001) compared with non-Hispanic White children. Hispanic children consumed snacks with lower energy density compared with non-Hispanic White children (β = −.20 [.05]; P < 0.001). However, the percent of daily energy from snacking was not associated with child race and ethnicity. Children of HH with high school or some college education consumed more snacks per day (β = .13 [.05]; P < 0.05), snacks with higher energy-density (β = .11 [.05]; P < 0.05), and a greater percent of daily energy from snacks (β = 1.41 [.58]; P < 0.05) compared with children of HH with less than a high school education.

Table 3.

Associations among sociodemographic variables and snack frequency, snack size, and snack energy density among 3,313 US children aged 2 to 5 years participating in the National Health and Nutrition Examination Survey 2007–2018 with at least one snacking occasion^a

	Snack frequency (occasion/person/day)	Snack size (kcal/occasion)	Snack energy-density (kcal/g/occasion)b	Percent of daily energy from snacks
Predictor	β (SE) 95% CI
Child age	.13 (.02)	22.44 (1.64)	.04 (.02)	1.64 (.20)
	(.09 to .17)***	(19.23 to 25.65)***	(−.01 to .08)	(1.23 to 2.05)***
Child gender
Femalec	–	–	–
Male	.34 (.05)	29.53 (3.91)	.00 (.04)	2.67 (.44)
	(.25 to 0.43)***	(21.75 to 37.30)***	(−0.08 to 0.08)	(1.80 to 3.55)***
Child race and ethnicity
Non-Hispanic Whitec	–	–	–
Non-Hispanic Black	−.16 (.05)	18.36 (4.80)	.08 (.06)	−.51 (.53)
	(−0.25 to −.07)**	(8.95 to 27.77)***	(−.04 to .19)	(−1.57 to .55)
Hispanic	−.02 (.06)	3.58 (4.45)	−.20 (.05)	.02 (.57)
	(−.14 to .09)	(−5.15 to 12.31)	(−.30 to −.10)***	(−1.12 to 1.15)
Other	.05 (.08)	13.45 (12.92)	.06 (.08)	.08 (.77)
	(−.10 to .19)	(−11.87 to 38.76)	(−.10 to .22)	(−1.45 to 1.61)
Head of household education
Less than high schoolc	–	–	–
High school/some college	.13 (.05)	5.51 (4.51)	.11 (.05)	1.41 (.58)
	(.03 to .24)*	(−3.32 to 14.35)	(.01 to .22)*	(0.26 to 2.56)*
College graduate or above	.09 (.07)	−.23 (5.83)	.06 (.07)	0.32 (.73)
	(−.04 to .23)	(−11.64 to 11.19)	(−.08 to .19)	(−1.13 to 1.77)

^aData were derived from demographic and What We Eat in America day 1 and 2 dietary intake files, National Health and Nutrition Examination Survey 2007–2018, weighted.^32–37

^bBased on all foods and beverages consumed as snacks.

^cReference category.

^*P < 0.05.

^**P < 0.01.

^***P < 0.001.

DISCUSSION

Current scientific knowledge of the associations of key snacking parameters (frequency, size, and energy density) with weight status among young children is limited. This research evaluated associations of snacking and weight among US preschool-aged children using NHANES data across the most recent decade for which dietary data are available (ie, 2007–2018). The findings reveal that US preschool-aged children (aged 2 to 5 years) with OW and OB snack more frequently, consume more calories per snacking occasion, and a greater percent of daily calories from snacking than children with NW. The association of snack energy density with child weight status was not statistically significant. These findings establish that associations of snacking frequency and size with OW and OB among children are evident as early as the preschool years.

The observed association of greater snacking frequency with higher weight status among US preschool-aged children agrees with the previous analysis of snacking among US children aged 1 to 5 years.²⁵ The present findings extend that work by revealing snack size (ie, calories per snack) as an important correlate of weight status among US preschool-aged children. In this analysis, the mean energy consumed per snack occasion was ~26 kcal higher among children with OW/OB than children with NW. This difference in mean snack size translates to approximately 75 kcal per day, when considering that children with OW/OB were reported to have, on average, 2.8 snacking occasions daily. The association of larger snack sizes (kcal/occasion) and higher weight status in this study is also consistent with associations reported in older children participating in NHANES. In particular, among children aged 6 to 19 years participating in 2003–2012 NHANES, greater snacking frequency was associated with increased risks of OW and OB among children aged 6 to 11 years only.³⁸ In addition, US adolescents aged 12 to 19 years participating in 2005–2016 NHANES with OW and OB consumed more frequent snacks daily and more calories per snacking occasion compared with those with NW.²⁷ Collectively, these studies provide population-representative cross-sectional evidence that frequent snacking and greater snack energy (per snacking occasion and as percent of daily calories) are associated with OW and OB among US preschool-aged children. Additional research is needed to evaluate causal influences of snacking on excessive energy intake among young children.

In this analysis, the energy density of snacks consumed by preschool-aged children in the US was not appreciably associated with weight status, regardless of the definition of snack energy density employed. These results are divergent with a small number of previous studies showing links between greater consumption of energy dense snack foods (eg, sweets, sugar-sweetened beverages, and crackers/chips) and higher weight status among young children.^57,58 For instance, a study of 432 preschool-aged children and their mothers with OB, reported that child dietary patterns characterized by processed snack type foods, were associated with increased odds of OB among children.⁵⁷ Another recent systematic review of 25 studies found an association of higher consumption of ultraprocessed foods, including snack foods, with higher body fat during childhood and adolescence.⁵⁸ Why the association between snack energy density and weight status of US children was not statistically significant in the present analysis is not immediately apparent but could reflect the definitions employed. For instance, it is possible that the characterization of specific types of energy-dense, nutrient poor snack foods (eg, ultraprocessed⁵⁸) in previous work is more indicative of dietary patterns that contribute to the development of obesity than the characterization of snack energy density in this research, which does not take into consideration the nutrient density of foods. For example, some energy-dense foods, such as cheese and nut butters, are also nutrient dense and can be included as part of a healthful diet to help young children meet nutrient needs for growth. Future research should seek to disentangle influences of snacking parameters on energy intake through experimental designs that permit causal inferences.

Demographic correlates of snacking patterns among young children are not well characterized. In the present study, the strongest sociodemographic correlates of snacking were age and gender where older children and male children consumed more snacks per day, greater calories per snacking occasion, and a greater percentage of daily energy from snacking than younger and female children, respectively. The extent to which these associations reflect age- and gender-related differences in eating behaviors vs energy requirements is unclear and merits additional consideration. In addition, non-Hispanic Black children consumed fewer snacks per day but more calories per snacking occasion relative to non-Hispanic White children. These differences in snack patterning did not translate to race and ethnicity differences in the percent of daily energy from snacking. Children from households reporting high school or some college education showed more frequent snacking, consumed more energy-dense snacks, and a greater percent of daily energy from snacking compared with children from households reporting less than high school education. These findings contrast a previous analysis of snacking among US children from 1977 to 2014 by Dunford and Popkin in which non-Hispanic Black children showed the greatest increase in intake of foods as snacks and children from households with lower education levels showed the greatest increases in energy consumed from snacks during the study period.¹⁰ The present findings are also somewhat parallel to those of a recent review of disparities in nutrition intake during childhood and adolescence, in which lower overall diet quality was observed among older children and adolescents and non-Hispanic Black children and adolescents.⁵⁹ Collectively, these findings highlight that snacking patterns among children vary by sociodemographic characteristics and point to populations where efforts to address excessive snacking may be particularly influential. Future research should seek to understand sociodemographic differences through a social determinants lens to provide insight into the broader social and economic environments in which snacking occurs. The significance of such research is highlighted by observations of targeted food marketing promoting fast-food, sugar-sweetened beverages, candy, and unhealthy snack brands to Black and Hispanic communities.¹⁹

Findings are qualified by several notable strengths and limitations. The simultaneous consideration of multiple snacking parameters of frequency, size, and energy density is a strength of the approach that advances conceptualization of snacking beyond frequency or generic categories of food types (eg, salty snacks).^8,25,38 Further, NHANES is the largest, longest-running, and most comprehensive survey of nutritional status and health of the US population and employs rigorous methods for collecting dietary recalls. However, misreporting remains a significant limitation for proxy-reported dietary recalls in children, influenced by memory recall, portion size estimation, and social desirability.^40,60 To address this limitation without biasing the sample, the present analyses were adjusted for energy misreporting by including the EI:EER.⁴⁰ Whether or not 2 days worth of intake represents typical consumption should also be considered when interpreting findings. The cross-sectional nature of the data does not permit causal inferences; longitudinal and experimental designs are needed to evaluate temporal ordering and causality.

CONCLUSIONS

The findings of this research provide population-representative evidence that US preschool-aged children with OW and OB consume more frequent snacks, more calories per snacking occasion, and a greater percent of daily energy from snacking than children with NW. The mechanisms through which snacking may influence weight status are not well understood. Whether or not snacking has a direct causal influence on excessive energy intakes among young children has not been established and is an important question for future research.

RESEARCH SNAPSHOT.

Research Question:

Are snack frequency, size, energy density, and percentage of daily energy associated with weight status among US preschool-aged children?

Key Findings:

In this cross-sectional analysis of 2007–2018 National Health and Nutrition Examination Survey data, US preschool-aged children aged 2 to 5 years with overweight/obesity consumed more frequent (occasions/day) and more calories per snack (kilocalories/occasion) as well as a greater percentage of daily energy from snacking than children with normal weight. The association of snack energy density with weight status was not statistically significant. Mean snack frequency and size as well as percentage of daily energy from snacking varied with child age, gender, and head of household education.