Factors correlated with ultra-processed food (UPF) intake in preschool-aged children and association with weight

Abstract

Objective:

Understand correlates of ultra-processed food (UPF) intake, examine association of UPF on BMI in 3-5-year-old children.

Design:

Secondary analysis of prospective cohort of 3-5-year-olds/parent, followed one-year between March 2014-October 2016. Usual UPF intake from two, 3-day food records completed one-year apart, a standardized nutrient database customized with child-specific foods, and NOVA food classification system was used. Child/parent characteristics and media use measured via parent-reported surveys. Child weight/height objectively measured.

Setting:

New Hampshire community.

Participants:

667 parent-child dyads screened, 624 enrolled with 90% follow-up.

Main Outcome Measure(s):

Primary outcome: identify correlates of UPF intake. Secondary outcome: determine if UPF intake associated with BMI change.

Analysis:

Adjusted beta linear regression, linear regression, P<0.05.

Results:

UPF accounted for 67.6% of total caloric intake. In adjusted models, children’s UPF intake was positively associated with increasing child age, greater hours watching TV, more frequent parent soda/fast food intake. UPF intake was negatively associated with higher parent education and reported race/ethnicity other than non-Hispanic White. No association for UPF intake and weight.

Conclusions and Implications:

There are several predictors of UPF intake in young children. Family-level interventions could be implemented to encourage intake of minimally processed foods before and during preschool years.

INTRODUCTION

Ultra-Processed Food intake

Ultra-Processed Foods (UPF) account for nearly 63% of calories purchased in US households¹, and data from the National Health and Nutrition Examination Survey (NHANES) shows that 58.2% of 2-5 year old’s diet is from UPF². Ultra-Processed Foods are industrial formulations made from substances derived from foods, with little intact food (e.g., sugar-sweetened beverages, savory packaged snacks)³. There are several classification systems for UPF, however, the NOVA (not an acronym) food classification system is widely used for health outcomes⁴. This system was developed by researchers in Brazil and provides a framework to group foods based on extent of food processing³. Ultra-Processed Foods are easy to over-consume as they are highly convenient, hyperpalatable, often low-cost and have a long shelf life¹. Overconsumption of UPF is associated with cardiometabolic conditions in children and adults^5,6 all-cause mortality and an increased risk of cancer in adults^7-11. Also, in adults and children some studies, but not all, link UPF intake with overweight and obesity^12,13. In a simulation model, reducing UPF intake by approximately 17% was associated with an estimated 16% and 9% reduction in obesity and overweight, respectively, in a nationally representative sample of children and adolescents¹⁴.

Impact of factors on total UPF intake in preschool-aged children

There are gaps in our understanding of factors related to UPF intake in preschool-age children. For example, few studies in preschool-aged children have focused on the relationship between certain household or individual factors (e.g., appetitive behaviors, physical activity, parent characteristics, rural/urban residence) and total UPF intake^15-18. One of these studies found a positive association between screen time and UPF intake¹⁶ and another study found no association¹⁷. Knowledge about these factors may help to inform interventions to reduce young children’s overall UPF intake.

Exposure to food marketing is a potential influence on children’s UPF intake. Examples of the types of foods marketed to young children include foods from fast food restaurants, yogurt, cereal, fruit snacks, sugar-sweetened beverages and savory packaged snacks¹⁹, foods that are mostly if not all, considered to be UPF. Several studies have found links with advertisement exposure and specific types of UPF (e.g., cereal, fast food)²⁰. For example, in our previous work, we showed that child-directed cereal TV advertising was prospectively associated with brand-specific cereal intake among preschoolers and separately, that advertisement exposure increased the risk of fast food intake among children when parents consumed fast food less frequently^21,22. Furthermore, research shows TV viewing may increase snacking, intake of certain processed foods, purchase requests for advertised foods and ultimately intake of foods that may be highly processed, however, to our knowledge, no study has examined food advertisement exposure and total UPF intake specifically, in young children²³. Additionally, while some studies have found a link with screen time and UPF intake, there is a lack of data on screen time with advertising exposure specifically^15-18. These gaps are important to address to understand how the marketing of specific foods may generalize to similar, highly palatable foods for young children.

UPF and association with overweight and obesity in children

Research suggests that the remainder of the diet aside from fast food (i.e., a Western dietary pattern not including fast food) is more strongly associated with obesity in children than fast food²⁴. To our knowledge, 4 studies including children aged 4 and 5 years old, reported mixed results for the association between UPF intake and weight^15-18. All 4 of these studies used the NOVA food classification system³, a common method to classify UPF, and collected dietary data through a variety of methods (i.e., 1-2 24-hour recalls^15,17, a Food Frequency Questionnaire¹⁶, 2-3 day food records¹⁸). One of these studies reported an association between higher UPF intake and increased weight in children¹⁸ but 3 other studies were null^15-17. These studies had several methodologic limitations, including use of cross-sectional designs^16,17, inability to control for socioeconomic status¹⁶, retrospective or limited diet collection methods^15-17, use of self-reported height and weight^16,17.

The objectives of this secondary analysis are: 1) to understand the potential correlates of UPF intake, such as media use, food advertising and other household or sociodemographic factors, and 2) to examine the association of UPF intake on obesity and overweight in 3-5-year-old children. We leveraged data from a prospective study with 2, 3-day food records and 2 objectively measured height and weight measurements each completed approximately 1-year apart, a widely-used food classification system⁴ (i.e., NOVA food classification system³) with a standardized nutrient database (Nutrition Data System for Research^25-27) customized to include child-specific packaged foods, and survey data including media use, extensive parent-child sociodemographic and household characteristics. Findings from this investigation can inform interventions and recommendations to reduce UPF intake among young children.

METHODS

Study design, participants, and recruitment

This study is a secondary analysis of data collected from a community-based prospective cohort of 624 3-to-5-year-olds and 1 of their parents, recruited in New Hampshire and enrolled between March 2014 – October 2016. The parent study investigated exposure to TV food marketing and intake of advertised cereals²¹ and fast food²², and fast food intake and weight change²⁸. Recruitment was conducted at pediatric outpatient clinics, federal assistance clinics, community and recreational events, and child care centers from 2 New Hampshire cities (Manchester and Nashua). Facebook and participant referrals were also accepted. Eligible children were: 3-5 years of age and living with a parent 3 or more days per week or every other week. Parents were also required to be English literate and lived within 1-hour drive from the recruitment sites. If multiple age-eligible children were present, 1 was randomly selected. Exclusion criteria included children with a health condition that impacted their food intake or the family had plans for relocation within a 1-year time frame.

At baseline, trained research assistants met with participants for an in-person visit which included: 1) measurement of parent and child height and weight and 2) a face-to-face training session for parents on how to complete and return a 3-day food record for their child. After the study visit, parents completed an online baseline survey covering a variety of factors, including child and parent media use, fast food and cereal intake, household behaviors (e.g., frequency of eating dinner together as a family), and sociodemographic information. Surveys were pre-tested cross-sectionally with a demographically comparable sample for comprehension, face validity and completion time^29,30. Every 2 months for approximately 1-year from baseline, parents were sent a shorter online survey (for information about media use, food intake (e.g., fast food, cereal), and supermarket use). Approximately 1-year from baseline, the parent-child dyads completed: 1) a final online survey (e.g., media use), 2) a second 3-day food record, and 3) repeated measures of height and weight for the child and parent.

Signed informed written consent was obtained from all parents (child assent was not required). Parents received $150 in gift cards if they completed all baseline and 1-year study components and children received 2 toys. The Dartmouth College Committee for the Protection of Human Subjects approved the study with full board review.

Dietary assessment

Parents were asked to record all food, beverages, and supplements, including brand name, for their child on 2 weekdays and 1 weekend day, 2 times approximately 1-year apart. Methods are detailed in previous analyses³¹ but briefly, all foods described in the food records were entered into the Nutrition Data System for Research (version 2014, University of Minnesota, Minneapolis, MN, 2014), a dietary assessment program developed by the Nutrition Coordinating Center (NCC) at the University of Minnesota. We customized the database with child-specific packaged foods that were not in the system (e.g., kid’s yogurt tubes, squeezable food pouches), food items which accounted for nearly 20% of the final dietary database. Trained research staff resolved confusing entries in the parent-reported record with the parent. For food record data and nutrient verification, data quality checks were conducted for a 5% subset of children, 15% of all days across all children, and 5% of the food records were randomly selected for double entry by an experienced NDSR analyst at an external site to help ensure accuracy of the data entry (Tufts University School of Medicine, Boston, MA). We checked for implausible values for mean total caloric intake across the available days for each child and none were outside of a defined range³² of <500 calories per day or >3500 calories per day (i.e., 724 to 3,167 calories per day.) We combined baseline and 1-year data for each child to represent usual total mean intake from UPF, to obtain a larger variation in food intake (i.e., ‘usual intake’) over the course of the year.

NOVA group assignment for UPF intake.

For each food item entered into the nutrient database across the study year (n=50,433), a NOVA food classification group 1-4 was assigned, with 1 as unprocessed food and 4 as ultra-processed food. Classifications of foods depend on the nature, purpose and extent of food processing and are categorized into 4 distinct groups. Group 1 is unprocessed and minimally processed foods including fruit, vegetables, meat, eggs, pasta, and water. Group 2 is processed culinary ingredients derived from Group 1 and includes items used to cook and season foods (e.g., salt, honey, sugar, butter). Group 3 are processed foods that have been changed to increase the durability or modify the palatability of unprocessed foods (e.g., foods with added sugar, salt, oil) such as canned fruit, cheese, and fresh bread. Group 4 are ultra-processed foods which are industrial formulations made from substances derived from foods, with little intact food (e.g., chips, chicken nuggets, cereal). It is important to note that the presence of at least 1 ingredient in a food product that is ultra-processed is sufficient to classify a product as an UPF³. This classification method is widely used in other studies examining UPF intake and health outcomes in children and adults^6,15-18.

Most food items included brand names per study protocol and these foods were assigned to NOVA groups based on ingredient lists found on food manufacturer’s websites. One limitation is that we did not calculate interrater reliability estimates as we had very detailed food record information with brand names. We checked for ingredients that were considered to be classified as ultra-processed, such as those found in Meadows et al., Supplementary Table S1³³. Unbranded food item ingredients were verified with the US Department of Agriculture (USDA) FoodData Central³⁴. All foods were conservatively estimated as the lowest processing classification applicable. To estimate the proportion of a child’s usual diet from UPF, we combined baseline and 1-year data for each child and computed the total calories from Group 4 (UPF) divided by the total calories of all days combined.

Media use and food advertising

To estimate overall time spent watching TV, we averaged estimates from baseline and 1-year survey. To categorize time spent watching TV further, from the baseline survey, media type (including advertisement-supported media [a proxy for food ad exposure], ad-free media, and mixed-ad media) had already been estimated for other analyses and details can be found there³¹. Briefly, at baseline the average hours per week for each of 6 types of media (TV, DVD/VHS, streaming, using apps, internet, game consoles and computer) were estimated. TV time also included individual networks viewed, which were classified as ad-supported (e.g., Nickelodeon) or ad-free (e.g., The Disney Channel). To estimate hours per week of ad-supported media use, we used an estimate of time spent on children’s TV channels that contained advertisements. To estimate hours per week of ad-free media viewing, we used an estimate of time spent on children’s TV channels that did not contain advertisements and DVD use. To estimate mixed-ad media, we combined estimates of time spent on the remaining types of media that may or may not contain advertisements (i.e., apps, internet, game consoles, computer). We did not capture actual UPF advertisements on the children’s TV networks, but we did confirm the presence of fast food and cereal advertisements on these networks^21,22. These estimates were previously used at baseline only³¹; we repeated the methods for the 1-year final survey. We averaged the measures from baseline and 1-year to obtain mean children’s total screen time (hours per week) over the study year, classified as ad-supported, advertisement-free, mixed-ad media use. Differences in screen time from baseline to 1-year were examined (Table 3) and the increase in total screen time hours is mostly from an increase in mixed-ad media.

TABLE 3.

CHILD SCREEN TIME AND AD EXPOSURE AND UNADJUSTED ASSOCIATIONS OF MEAN UPF¹ INTAKE OVER THE STUDY YEAR AMONG PRESCHOOL AGED CHILDREN (N=560).

	n (%)	UPF1 Intake Mean (SD)	P-value3
Mean total screen time	23.62 (15.1)	r = 0.15	<0.001
Baseline	17.61 (14.2)	r = 0.17	<0.001
Year-one	29.63 (20.4)	r = 0.10	0.02
Mean TV time only
As continuous, mean (SD)
Hours per week	8.11 (6.5)	r = 0.17	<0.001
As ordinal, n (%)
0 hours per week	68 (12.1)	0.63 (0.17)	<0.001
1-7 hours per week	229 (40.9)	0.67 (0.11)
8-14 hours per week	180 (32.1)	0.70 (0.11)
14-21 hours per week	65 (11.6)	0.70 (0.11)
≥21 hours per week	18 (3.2)	0.69 (0.10)
Screen time stratified by ad exposure 2
Ad-supported TV
As continuous, mean (SD)
Hours per week	0.88 (1.59)	r = 0.14	0.001
As ordinal, n (%)
0 hours per week	317 (68.6%)	0.66 (0.13)	0.008
1-7 hours per week	238 (30.3%)	0.69 (0.11)
8-14 hours per week	5 (1.1%)	0.70 (0.11)
Ad-free TV, DVD or VHS
As continuous, mean (SD)
Hours per week	6.64 (5.41)	r = 0.19	<0.001
As ordinal, n (%)
0 hours per week	46 (8.5%)	0.61 (0.19)	<0.001
1-7 hours per week	293 (50.5%)	0.67 (0.11)
8-14 hours per week	165 (30.6%)	0.70 (0.11)
14-21 hours per week	46 (8.5%)	0.72 (0.10)
≥21 hours per week	10 (1.9%)	0.69 (0.12)
Mixed-ad media
As continuous, mean (SD)
Hours per week	10.92 (10.4)	r = 0.05	0.27
As ordinal, n (%)
0 hours per week	27 (5.0%)	0.68 (0.14)	0.36
1-7 hours per week	219 (37.0%)	0.66 (0.12)
8-14 hours per week	173 (32.0%)	0.68 (0.13)
14-21 hours per week	65 (12.0%)	0.70 (0.13)
≥21 hours per week	76 (14.0%)	0.68 (0.12)

Among 560 parent-child dyads enrolled in a cohort study who completed a food record.

Entries that do not sum to 560 exclude missing data.

¹Ultra-Processed Food.

²Ad-supported TV included viewing of any of the six predefined ad-supported children’s TV networks (Boomerang, Cartoon Network, Disney XD, the HUB, Nickelodeon, or Nicktoons). Ad-free TV included viewing of any of the five predefined ad-free children’s TV networks (Disney Channel, Disney Jr., Nick Jr., PBS Kids or Sprout) and was combined with DVD or VHS viewing. Mixed-ad media included mobile applications, Internet and video or computer games.

³P-values are from one-way ANOVAs or Pearson Correlation Coefficients (r), P<0.05 for significance.

Other potential correlates of UPF intake

In the baseline survey, information on parent-reported items for child characteristics [age, sex (i.e., boy, girl), race, ethnicity, BMI, enrollment in childcare or school], child health behaviors (hours of outdoor play, hours of nighttime sleep, TV in the child’s bedroom, and usual fast food intake), parent and household characteristics (parent age, educational attainment, annual household income, WIC (Special Supplemental Nutrition Program for Women, Infants, and Children) participation, lives with spouse or partner, and number of household children), parent diet indicators (fruit and vegetable intake, soda intake, fast food intake) and family eating habits (number of days per week eating dinner together) were evaluated to determine potential associations with UPF intake. Race was self-reported by the parents of the children from a list including White, Black or African American, American Indian or Alaska Native, Asian, Native Hawaiian or Pacific Islander, Other (specify) and could select multiple options (i.e., multiracial). We asked if the child was Spanish, Hispanic, or Latino (Yes, No) for information on ethnicity.

BMI difference from baseline and 1-year

Children’s height and weight were collected at both baseline and 1-year visits by trained research assistants using a standardized protocol^35,36. Height was measured in centimeters with a study stadiometer and repeated 2 times. Weight was measured in kilograms to the nearest 0.1 kg on a study scale that was accurately calibrated (SECA brand) and repeated 2 times. To estimate BMI difference from baseline to 1-year, we used age-and-sex adjusted BMI (Body Mass Index) percentiles, as provided by formulas from the Center for Disease Control and Prevention (CDC)³⁷, with overweight defined as an age- and sex-adjusted BMI ≥ 85^th and <95^th percentile, obesity defined as an age- and sex-adjusted BMI ≥ 95^th percentile, and healthy weight defined as an age- and sex-adjusted BMI <85^th percentile. There were very few children in the underweight category (n=10) and they were included in the healthy weight category for this analysis. For BMI difference across the study year, the age- and sex-adjusted BMI percentiles from baseline were subtracted from the 1-year BMI percentiles. For secondary outcomes, we also evaluated baseline BMI and for a subset of children who moved from healthy BMI percentiles at baseline to overweight or obese BMI percentiles at 1-year.

Statistical analyses

Potential correlates of UPF intake were summarized overall and by UPF intake. To examine adjusted associations of correlates on UPF intake, beta regression was used to account for using a proportion as an outcome and beta coefficients are reported on the logit scale. Model assumptions were evaluated using residual plots and determined to be met. Similarly, potential correlates for BMI difference were summarized. To examine adjusted associations of correlates on BMI difference, linear regression was used. Model assumptions were determined to be met using statistical tests (e.g., Variance Inflation Factor) and looking at residual plots. For secondary analyses for BMI, we used baseline BMI in linear regression models and separately, we took a subset of children who transitioned up a BMI category and used this subset in an adjusted linear regression model. Overall, model predictors included measures associated with either UPF intake or BMI difference at the P<0.10 level from bivariate analyses. Main effects at the P<0.05 level were considered statistically significant in adjusted analyses. Analyses were completed with the R Language for Statistical Computing (version 4.0.2, R Foundation for Statistical Computing Platform, https://www.r-project.org, 2022).

RESULTS

Overall, 667 children were screened and 624 enrolled (93.6%). A total of 560 parent-child dyads were included in the analysis for potential correlates associated with UPF intake, as they completed the baseline survey and had at least 1 food record completed. In this analytic sample, approximately 99% had 3 or more days of daily food recorded at either timepoint (84% have all 6 days completed). Similar to other studies with this sample^28,31, those missing data (of the 624 enrolled) were more likely to include children who were boys, were non-White, Hispanic, had parents of lower education with lower household incomes, had parents more likely to live with a spouse or partner, had greater hours of watching TV, had more ad-free and mixed-ad media use, and were more likely to have a TV in their bedroom (P≤0.01). There was no difference for age, overall screentime, and ad-supported media use (P>0.05) (data not shown). Furthermore, a total of 522 parent-child dyads were included in the analysis for UPF intake associated with BMI difference, as they completed the baseline survey and had height and weight data at the 1-year follow-up.

Table 1 presents counts, proportions, and unadjusted associations of child characteristics and behaviors and UPF intake. The analytic sample included mostly 3-year-olds, slightly more girls, and children were mostly White, non-Hispanic. Over 1/3 of children reported greater than 21 hours of outside play each week and most children got at least 10 hours of sleep per night. Mean UPF intake was approximately 67.6% of children’s usual diet, with more than 93% of the children consuming at least 50% of their diet from UPF. The estimated mean intake from children’s diet for minimally processed foods (Group 1) was 25.8%, culinary ingredients (Group 2) was 1.7% and processed foods (Group 3) was 6.4%. Additionally, the mean total energy per day was approximately 1,446 (SD: 305) calories. In unadjusted analyses, a higher UPF intake was associated with increasing child age, having a TV in the child’s bedroom, and more frequent fast food intake, while a lower UPF intake was associated with a race or ethnicity other than non-Hispanic White.

TABLE 1.

CHILD CHARACTERISTICS AND BEHAVIORS AND UNADJUSTED ASSOCIATIONS OF MEAN UPF¹ INTAKE OVER THE STUDY YEAR AMONG PRESCHOOL-AGED CHILDREN (N=560).

	N (%)	UPF1 Intake Mean (SD)	P-value2
Overall	560 (100%)	0.68 (0.12)	--
Child characteristics
Age
3 years	227 (40.5%)	0.65 (0.13)	<0.001
4 years	211 (37.7%)	0.69 (0.12)
5 years	122 (21.8%)	0.71 (0.10)
Sex
Girl	313 (55.9%)	0.67 (0.11)	0.48
Boy	247 (44.1%)	0.68 (0.13)
Race / ethnicity
Non-Hispanic White	483 (86.5%)	0.68 (0.12)	0.02
Other	77 (13.8%)	0.65 (0.15)
Weight status (baseline)
Healthy weight	405 (72.3%)	0.67 (0.12)	0.25
Overweight	95 (17.0%)	0.69 (0.12)
Obese	60 (10.7%)	0.67 (0.12)
Childcare outside of the home
None	106 (18.9%)	0.66 (0.14)	0.26
1 - 20 hours per week	194 (34.6%)	0.69 (0.13)
21+ hours per week	260 (46.4%)	0.67 (0.11)
Child health behaviors
Hours of outside play
<14 hours per week	213 (38.0%)	0.67 (0.13)	0.76
14 to 21 hours per week	143 (25.5%)	0.68 (0.11)
>21 hours per week	204 (36.4%)	0.68 (0.12)
Nighttime sleep
<10 hours per night	44 (8.2%)	0.68 (0.12)	0.90
≥10 hours per night	491 (91.8%)	0.68 (0.12)
TV in the bedroom
Yes	103 (18.7%)	0.72 (0.11)	<0.001
No	448 (81.3%)	0.67 (0.12)
Usual fast food intake
Less than once a month	196 (35.0%)	0.63 (0.13)	<0.001
At least monthly to less than once a week	212 (37.9%)	0.69 (0.12)
At least weekly	152 (27.1%)	0.72 (0.11)

Among 560 parent-child dyads enrolled in a cohort study who completed a food record.

Entries that do not sum to 560 exclude missing data.

¹Ultra-Processed Food.

²P-values are from one-way ANOVAs or two-sample t-tests, P<0.05 for significance.

Table 2 presents counts, proportions, and unadjusted associations of parent and household characteristics and UPF intake. Most parents had a bachelor’s degree or higher and a higher annual household income. Slightly more than 1/3 of parents reported roughly meeting the USDA dietary guidelines of 2, 1-cup servings of fruit and 2 1/2, 1-cup servings of vegetables per day (for a diet of 2000 calories)³⁸. Additionally, nearly half of parents reported eating dinner together as a family every day of the week. In unadjusted analyses, UPF intake was positively associated with estimates of parent diet, including more frequent fast food intake, higher soda intake, lower intake of fruits and vegetables, and eating dinner together as a family 7 days a week. Also, a lower UPF intake was associated with increasing parent age, higher parental education, higher household income, and a parent residing with a spouse or partner.

TABLE 2.

PARENT AND HOUSEHOLD CHARACTERISTICS AND UNADJUSTED ASSOCIATIONS OF MEAN UPF¹ INTAKE OVER THE STUDY YEAR AMONG PRESCHOOL-AGED CHILDREN (N=560).

	n (%)	UPF1 Intake Mean (SD)	P-value3
Parent characteristics
Age
20-29 years	108 (19.3%)	0.70 (0.12)	0.05
30-39 years	363 (64.9%)	0.68 (0.12)
40 years and older	88 (15.7%)	0.65 (0.12)
Education
Up to an Associate’s or technical degree	220 (39.3%)	0.70 (0.11)	<0.001
Bachelor’s degree or higher	340 (60.7%)	0.66 (0.13)
Annual household income
<$75,000	234 (41.8%)	0.69 (0.12)	0.04
≥$75,000	326 (58.2%)	0.67 (0.12)
Receives WIC benefits
Yes	69 (12.3%)	0.69 (0.13)	0.24
No	491 (87.7%)	0.67 (0.12)
Parent co-habitation status
Lives with spouse or partner	482 (86.4%)	0.67 (0.12)	0.02
Does not live with spouse or partner	76 (13.6%)	0.71 (0.11)
Parent diet indicators
Daily vegetable and fruit intake2
Below guidelines for both	206 (36.9%)	0.70 (0.12)	0.002
Below guidelines for one	141 (25.3%)	0.67 (0.12)
Meets guidelines for both	211 (37.8%)	0.66 (0.12)
Frequency of soda intake
0 days a week	409 (73.8%)	0.66 (0.13)	<0.001
1 to 2 days a week	76 (13.7%)	0.71 (0.11)
3 or more days a week	69 (12.5%)	0.71 (0.11)
Frequency of fast food intake
Less than once a month	178 (31.8%)	0.64 (0.13)	<0.001
At least monthly yet less than once a week	164 (29.3%)	0.69 (0.12)
At least weekly	218 (38.9%)	0.70 (0.11)
Eat dinner together as a family, days per week
4 days or less	134 (24.0%)	0.70 (0.12)	<0.001
5 or 6 days	161 (28.8%)	0.68 (0.13)
7 days	264 (47.2%)	0.66 (0.12)
Household characteristics
Number of children in the home
1	120 (21.4%)	0.68 (0.12)	0.26
2	300 (53.6%)	0.67 (0.12)
3	96 (17.1%)	0.69 (0.12)
4 or more	44 (7.9%)	0.71 (0.13)
Household number
2	84 (16.9%)	0.67 (0.13)	0.15
3	270 (54.2%)	0.67 (0.12)
4	98 (19.7%)	0.68 (0.12)
5	40 (8.0%)	0.72 (0.12)
6	6 (1.2%)	0.64 (0.09)

Among 560 parent-child dyads enrolled in a cohort study who completed a food record.

Entries that do not sum to 560 exclude missing data.

¹Ultra-Processed Food.

²Below guidelines was defined as less than 2–3 servings of fruit or vegetables, separately, per day while meets guidelines was defined as at least 2–3 servings per day.

³P-values are from one-way ANOVAs or two-sample t-tests, P<0.05 for significance.

Table 3 presents child screen time, time watching TV only, ad exposure and unadjusted associations between UPF intake and screentime. Children averaged 23.6 (SD: 15.1) hours of total screen time per week across the study year, with 8.11 (SD: 6.5) hours per week from TV only. Across the study year, mean viewing of ad-supported media use accounted for the smallest portion of all media use, with a mean of 0.88 (SD: 1.6) hour per week and children’s ad-free media had a mean of 6.64 (SD: 5.4) hours per week. Children’s mixed-ad media accounted for the greatest share of all media use, averaging 10.9 (SD: 10.4) hours per week. In unadjusted analyses, UPF intake was positively associated with increased media use, more time spent watching TV, increased viewing of ad-supported media and ad-free media. Furthermore, UPF intake was positively correlated with mean total screen time hours per week (P < 0.001), mean TV time only (P < 0.001), viewing ad-supported media (P = 0.001), and ad-free media (P < 0.001).

In the fully adjusted beta regression model shown in Table 4, we found older age to be associated with an increase in mean UPF intake (P ≤ 0.001). The beta coefficient ($\widehat{\beta }:0.13$) translates into a 14% increase in mean UPF intake (as % of total calories) for each 1-year increase in age because the dependent variable was log-transformed. Furthermore, we found parents reporting child race or ethnicity other than non-Hispanic White to be associated with a lower mean UPF intake (P ≤ 0.01), where the beta coefficient translates into a 17% lower mean UPF intake (as % of total calories). Higher parent education level (P = 0.02), parent consumption of fast food reported as less than once a month compared to more often (P < 0.001), and parents reporting no soda intake compared to 1-2 days a week (P = 0.04) were also associated with a lower mean UPF intake, while an increase in mean weekly TV viewing was associated with a higher mean UPF intake (P = 0.01) (Table 4).

TABLE 4.

UNADJUSTED AND ADJUSTED ASSOCIATIONS BETWEEN POTENTIAL CORRELATES OF UPF¹ INTAKE AND MEAN UPF¹ INTAKE AT BASELINE AND ONE YEAR, AMONG PRESCHOOL AGED CHILDREN (N=560).

	Pseudo r2	Unadjusted3 b2 (95% CI)	Adjusted4 b (95% CI)	P-value5
Fully adjusted model 3	0.16
Child characteristics
Age (continuous)	0.03	0.13 (0.07, 0.19)	0.13 (0.07, 0.19)	<0.001
Child sex: girl vs. boy	0.00	−0.05 (−0.14, 0.05)	−0.05 (−0.14, 0.04)	0.28
Race, Ethnicity: Non-Hispanic White vs. Other	0.01	−0.16 (−0.29, −0.03)	−0.21 (−0.34, −0.08)	0.002
Mean TV (hours per week)	0.02	0.14 (0.01, 0.02)	0.01 (0.00, 0.02)	0.01
Parent characteristics
Parent Age	0.01
20-29 years		Ref	Ref
30-39 years		−0.11 (−0.23, 0.01)	−0.03 (−0.15, 0.10)	0.65
40 years and older		−0.19 (−0.35, −0.04)	−0.13 (−0.29, 0.03)	0.11
Parent Education Level	0.04
Up to an Associate’s or technical degree		Ref	Ref
Bachelor’s degree or higher		−0.22 (−0.32, −0.12)	−0.12 (−0.23, −0.02)	0.02
Household income: ≥$75k vs. <$75k	0.01	0.12 (0.02, 0.21)	0.01 (−0.10, 0.11)	0.92
Parent lives with spouse or partner	0.01	−0.17 (−0.31, −0.03)	−0.09 (−0.24, 0.05)	0.20
Parent diet indicators
Parent vegetable and fruit intake	0.02
Meets guidelines for both		Ref	Ref
Below guidelines for one		0.03 (−0.09, 0.15)	−0.01 (−0.11, 0.12)	0.83
Below guidelines for both		0.18 (0.07, 0.29)	0.10 (−0.00, 0.21)	0.06
Parent soda intake per week
0 days a week	0.03	Ref	Ref
1 to 2 days a week		0.20 (0.06, 0.34)	0.14 (0.00, 0.27)	0.04
3 or more days a week		0.23 (0.08, 0.38)	0.06 (−0.09, 0.21)	0.45
Parent fast food intake	0.05
Less than once a month		Ref	Ref
At least monthly yet less than once a week		0.22 (0.10, 0.34)	0.21 (0.10, 0.32)	<0.001
At least weekly		0.30 (0.19, 0.41)	0.25 (0.14, 0.36)	<0.001
Eating dinner with family (days per week)	0.01
4 days or less		Ref	Ref
5-6 days		−0.10 (−0.23, 0.03)	−0.09 (−0.22, 0.03)	0.15
7 days		−0.16 (−0.27, −0.04)	−0.11 (−0.22, −0.01)	0.06

Among 560 parent-child dyads enrolled in a cohort study who completed a food record.

¹Ultra-Processed Food.

²b = Estimated beta regression beta coefficient.

³Beta coefficients are from an unadjusted beta regression model.

⁴Beta coefficients are from a fully adjusted beta regression model; all model predictors are presented.

⁵P-values are from adjusted beta regression models, P<0.05 for significance.

In adjusted models where overall screen time was stratified by ad exposure, we found hours of ad-supported media (i.e., children’s TV networks with ads) was not associated with mean UPF intake over the study year (P=0.15), however, increased hours of ad-free media (i.e., children’s TV networks without ads) was associated with UPF intake over the study year (P=0.01). Mixed-ad media use was not associated with UPF intake. As described above there was limited range in viewing hours for advertisement supported media.

Overall, BMI difference was not associated with mean UPF intake. When we regressed BMI difference on UPF intake, no covariates were statistically related to BMI difference except for younger parent age (Supplementary Table S1). For secondary analyses, both baseline BMI and the subset of children who moved from healthy weight to overweight or obese were not associated with mean UPF intake. No covariates were statistically related to BMI difference except for younger parent age.

In sensitivity analyses shown in Supplementary Table S2, we found no association between UPF intake and BMI difference when we adjusted for baseline BMI (Model 1). We also tested for an interaction between baseline BMI and UPF intake and found no significant interaction (Model 2). Furthermore, we evaluated screen time stratified by ad media type as “any versus none” compared to UPF intake and found similar results as the initial findings (data not shown). In additional analyses (Supplementary Table S3), we evaluated the association between age and unprocessed or minimally processed foods (Group 1) (as opposed to UPF, Group 4). We found an unadjusted association between increasing age and lower proportion of unprocessed food intake, the opposite direction we found for UPF intake.

DISCUSSION

In this study, children consumed high levels of UPF (67.6% of total calories), slightly higher than the NHANES estimate (58.2%)². In adjusted models, we found mean UPF intake to be positively associated with increasing child age, greater hours of watching TV, parents drinking soda 1-2 days per week compared to none, and parents consuming fast food at least monthly and at least weekly compared to less than once a month. We found mean UPF intake to be lower in children reporting a race or ethnicity other than non-Hispanic White and those with higher parent education. Identifying these factors are important to inform potential interventions and identify targets to reduce young children's UPF intake.

The high levels of UPF intake observed in this study are concerning. Studies show that children’s palate and eating patterns may be largely established by the preschool-years and may continue into adulthood^39,40. In this study, increasing child age, even within the limited range of 3-5 year olds, was associated with increasing UPF intake. Our data indicate that UPF are replacing healthier choices. Therefore, interventions or actions to support more consumption of unprocessed or minimally processed foods, before and during the preschool-years will be essential to help offset the intake of UPF.

We found several parent- and family-dietary related factors to be associated or marginally associated with mean UPF intake. Increased frequency of parent soda intake and parent fast food intake were important predictors of higher mean UPF intake in children. In this sample, over 60% of parents did not meet the recommended guidelines for fruit and vegetable intake⁴¹ and not meeting the guidelines for both fruits and vegetables was marginally associated with higher mean UPF intake. Additionally, eating dinner together as a family every day of the week is important; other studies have shown that older children and adolescents tend to eat healthier when eating together with their family^42,43 and in our study, eating dinner together with family 7 days a week was marginally associated with lower mean UPF intake in children. Overall, these factors support that a family-level intervention focused on family meal patterns and parent dietary intake may be an important target to reduce UPF intake among young children.

We also examined sociodemographic factors for the parent and higher parental education level was an important predictor. Two studies in preschool-aged children^15,18, found parent education level to be an important predictor for UPF intake (but did not share effect sizes), while 2 other studies in preschool-aged children did not find parent education to be a predictor of UPF intake^16,17. It is possible that a high parent education relates to greater nutritional knowledge, however, future studies are needed to specifically determine this relationship.

Additionally, our data show a race or ethnicity other than non-Hispanic White is associated with lower UPF intake, potentially suggesting non-White and Hispanic children may be eating foods that are more culturally tailored and homemade, may prepare more foods at home, or it may be indicative that their overall diet has not fully shifted to the traditional “western” diet^44-46. A few studies support that those with higher acculturation indexes also have higher UPF intake, and those that are foreign-born have a lower UPF intake compared to US-born^44-46. Specifically in our study, we noticed an outlier that had a remarkably low UPF intake and when analyzing their individual dietary data, we found that they consumed mostly homemade Indian dishes such as curries, dal, fresh fruits and vegetables. However, a limitation of the study is that the non-White or Hispanic sample was very small and includes a variety of racial and ethnic groups. One previous study in preschool-aged children found no difference in UPF intake between White and non-Whites¹⁶ and another study in 2-11 year olds found no difference in fast food intake between non-Hispanic White, non-Hispanic Blacks, and Hispanic origin groups, however with 12-19 year olds, non-Hispanic Black and Hispanic adolescents consumed a higher intake of fast food compared to non-Hispanic Whites⁴⁷. More research in young children is needed to understand race and ethnicity as a predictor of UPF intake.

The data also shows a greater time spent watching TV was associated with a higher mean UPF intake. Four studies^15-18 in preschool-aged children included screen time duration in their investigation of the relationship between UPF intake and BMI. One of these studies found screen time duration to be an important predictor for UPF intake¹⁶ and 2 studies included this variable in their analysis but did not share effect sizes or confidence intervals for this predictor^15,18, while a cross-sectional telephone survey study found no association between screen time duration and UPF intake¹⁷. Advertisement exposure specifically was not examined in these studies.

Furthermore, while greater time spent watching TV was associated with a higher mean UPF intake, our findings do not support our hypothesis that young children’s exposure to ad-supported media (i.e., children’s TV networks with ads), a proxy for food-advertisement exposure, results in a higher intake of UPF, independently of other correlates of UPF intake. A possible explanation for this is there may have been too little variability in this dataset to see if ad-supported media use is related to UPF intake, but a previous study using the same dataset we found ad-supported media to be associated with the Healthy Eating Index (HEI), a measure of diet quality³¹. A limitation of our study is that we did not capture actual UPF advertisements on these children’s TV networks, but we did confirm the presence of fast food and cereal advertisements on these networks. This could also be explained by foods that are ‘healthier’ may contribute to a higher HEI score but may still be highly processed. Another possible explanation is that we found ad-free media exposure and total time spent watching TV associated with UPF intake, suggesting overall media use is still important. This relationship could be confounded by unmeasured factors such as more relaxed household rules for child health behaviors and habits, including media use and eating behaviors.

Additionally, there are no national diet recommendations for UPF intake, despite the potential negative health impacts of these foods. Dietary recommendations may be forthcoming, thus, understanding factors that relate to UPF intake are important to investigate. For example, such recommendations may be informed with more research evaluating factors associated with UPF intake, that account for household factors, such as screentime and family size. If screentime is associated with UPF intake in preschoolers, effective interventions to educate families to reduce screen time for their children and encourage greater intake of fruits and vegetables may be developed.

Also, a previous study in Spanish preschool-aged children reported family size as a significant risk factor of total UPF intake in children¹⁶, however, we found both household size and number of household children were not associated with UPF intake, in analyses with and without adjustment for household income. The impact of family size on UPF intake may vary across different geographic and cultural regions.

As a separate outcome, we found intake of UPF was not statistically significantly associated with BMI difference from baseline to 1-year in the unadjusted and adjusted models. This finding may suggest that the consumption of UPF may take a longer amount of time than 1 year to accumulate extra adiposity in children or that adiposity is predicted by other factors aside from UPF intake (e.g., gene-environment interaction⁴⁸, physical activity intensity levels⁴⁹). Additionally, overall, there was very little variation in weight difference across the study year, and specifically, only 43 children increased in weight status over the study year (i.e., moved from healthy weight at baseline to overweight or obese at 1-year). This finding aligns with other studies involving children of slightly older ages. Two studies, both from Brazil, showed no association between UPF intake and weight prospectively in 4-8 year olds¹⁵ and cross-sectionally in 7-10 year olds⁵⁰. Another prospective study found an increase in BMI at age 10 based on UPF intake at age 4, but not at age 7 for later BMI¹⁸, suggesting a longer timeline for increased weight status. Prior studies that have included older children and adolescents^50-57 have reported associations between total UPF intake and risk of overweight and obesity. These studies find UPF intake was associated with a higher risk of overweight and obesity in 4 studies^51-54 but not in 4 other studies^15,50,56,57. Therefore, it is important to further investigate total UPF intake in 3-5 year old children and the potential effect on weight over time.

Strengths of this study include looking at data prospectively from baseline to 1-year, the study taking place in a natural setting, using 2, 3-day food records with a customized nutrient database with child-specific foods (e.g., squeezable food pouches and yogurt tubes), having brand-specific data to look up food manufacturer information for UPF ingredients, using a widely-used food classification system (NOVA)³, and having extensive set of variables. It is important to note that most dietary studies in children have not customized the nutrient database with child-specific foods, potentially leading to misclassification and error, as food items can vary in nutritional content depending on individual packaging⁵⁸. Another strength includes that in previous studies with this dataset, we had confirmed the presence of food-advertisements on the children’s TV networks in this study sample^21,22.

Study limitations include the food item manufacturer information were checked in 2022, and the foods were consumed in 2014-2016, so there may have been products that changed over time, however there is no reason to believe that children aged 3-5 today are no longer eating the foods the children ate in the recent past. The NOVA food classification system is also being debated, on the utility of using processing level versus nutrients for diet recommendations and research^59-61, however, this method is widely used among researchers. Parent self-report of their child's diet and the potential predictor variables, such as media use and eating habits, could be affected by social desirability bias, reporting bias, and recall error, were used to complete this analysis. Also, ad-supported media and mixed-ad media were used as a proxies for food advertisement exposure, however, other studies support this assumption^62-64. Another limitation is that exposure to other avenues of food marketing, such as in supermarkets and by other family members media use (e.g., TV or online videos that are on in the background at home), were not captured with the presented data.

Also, parent UPF intake was not measured in this study; instead, we looked at parent sugar-sweetened beverages, fast food, and fruit and vegetable intake as proxies for overall UPF intake. However, intake of these foods are directly classified as UPF themselves³ and reflect a proportion of parent diet. Overall, these results could be affected by residual confounding from unmeasured variables and causality cannot be confirmed. Finally, this sample is reflective of a largely non-Hispanic, White population (~86%), characteristic of New Hampshire communities, where recruitment occurred⁶⁵. Therefore, results may not be generalizable to the entire US; more research is needed to evaluate UPF intake in diverse populations.

IMPLICATIONS FOR RESEARCH AND PRACTICE

These findings suggest that children during the preschool ages are already consuming high amounts of UPF and there are several predictors of child UPF intake. In this analysis of 560 preschool-age children who were recruited from community settings, we found increasing child age, lower parent education, a race or ethnicity of White, non-Hispanic, higher weekly TV viewing, and frequent parent soda and fast food intake to be predictors of a higher mean UPF intake in children. While increased hours of TV viewing was a predictor of mean UPF intake, exposure to ad-supported media was not associated with a higher UPF intake across the study year, independent of other correlates of UPF intake. UPF intake was also not associated with BMI difference across the study year. Importantly, childhood is an important timeframe to shape and support intake of healthy foods (i.e., minimally or unprocessed foods), therefore, interventions at the family level to encourage intake of minimally and unprocessed foods before and during the preschool-age years are essential to offset the high UPF intake observed. Furthermore, the media landscape is changing to include more mobile media, apps, online videos and websites that potentially have embedded unhealthy food advertisements. Additional research is needed to understand how current food marketing exposure across newer media platforms influences children’s overall UPF intake and any potential health effects.