Associations between Snacking and Weight Status among Adolescents 12–19 Years in the United States

Snacking is a significant contributor to energy intake among adolescents, but its association with weight status is unclear. To elucidate this association, data from 6545 adolescents (12–19 years) in the 2005–2016 National Health and Nutrition Examination Survey (NHANES) were analyzed. The mean number of daily snack occasions, mean snack size, and mean snack energy density were examined by weight classification (body mass index (BMI)-for-age percentiles: normal weight (NW) <85th; overweight (OW) ≥85th to <95th; obese (OB) ≥95th). Models included all snacking parameters, mean meal size, demographic characteristics, survey cycle year, and dietary reporting accuracy. Adolescents with NW consumed fewer snacks daily (1.69 (0.02) snacks/day) and smaller snacks per occasion (262.32 (4.41) calories (kcal)/snack) compared to adolescents with OW (1.85 (0.05) snacks/day, p = 0.005; 305.41 (8.84) kcal/snack, p < 0.001), and OB (1.97 (0.05) snacks/day; 339.60 (10.12) kcal/snack, both p < 0.001). Adolescents with OW and OB also consumed more added sugar, saturated fat and sodium from snacks, but had lower mean energy density per snack compared to snacks consumed by NW adolescents. US adolescents with OW and OB consume more snacks daily and more calories at each snacking occasion compared to adolescents with NW. Future studies should examine the prospective associations between snacking and weight status and impact on overall diet quality.


Introduction
Snacking (i.e., eating in between meals) has significantly increased in the United States (US) over the past 30 years alongside rates of pediatric obesity [1,2]. Adolescents consume approximately 20% of their daily energy intake as snacks, with salty and sweet foods contributing a majority of the snacking calories [3,4]. Current Dietary Guidelines for Americans recommend that adolescents consume nutrient-dense snacks to help them meet their nutritional needs [5]. However, whether snacking currently benefits dietary intake and growth among adolescents or promotes obesity is unknown [6,7].
Given that adolescents have the highest prevalence of obesity among children, understanding the relationship between snacking and weight status in this population is critically important [2]. Adolescents with obesity are at increased risk of diet-related co-morbidities including hypertension, type 2 diabetes, and dyslipidemia as well as psychosocial consequences such as stress, stigma and depression [8,9]. Compared to other pediatric populations, adolescents with obesity are also more likely to remain obese as adults, increasing their risk for additional morbidity and mortality [10,11]. A myriad of foods have been targeted as key dietary drivers of obesity, including sugary drinks,

Demographics
Age (years), gender (male/female), and race/ethnicity (non-Hispanic white, Hispanic/Mexican American, non-Hispanic black, other) were collected for the participating adolescents from the in-home interview. Head of household (HH) age, marital status, education level, and income were also collected. The ratio of income to poverty (RIP) was estimated using family income relative to the poverty threshold; families greater than 125% of the poverty threshold were classified as high RIP and those less than 125% of the poverty threshold were classified as low RIP [26].

Weight Status
Adolescent height and weight were objectively measured by trained staff in the mobile examination center [27]. Height was measured to the nearest 0.1 cm using a stadiometer and weight was measured to the nearest 0.1 kg using a digital scale. Height and weight were used to calculate body mass index (BMI) and date of birth and gender were used to derive age-and-sex specific BMI-percentiles (BMI%) using the CDC growth charts [28]. Weight status was classified according to the following BMI% ranges: normal weight (NW) <85th; overweight (OW) ≥85th to <95th; and obesity (OB) ≥95th [29].

Dietary Intake
Dietary data were collected via two 24-h dietary recalls [30]. The first recall was conducted in person and the second occurred via phone. Adolescents (12 years and older) self-reported their dietary intake using the United States Department of Agriculture (USDA) Automated Multiple-Pass method [31,32]. Eating occasions were characterized by participants using a pre-determined list (e.g., "breakfast"). Snacking occasions were those identified by participants as "snacks." Additionally, given the general definition of snacking that refers to eating in between meals, those occasions labeled as "beverages" or "extended consumption" were also included. Only adolescents with two days of dietary data were included in the analyses (1720 participants were excluded due to having only one dietary recall). The mean of two 24-h dietary recalls was used to estimate mean daily snacking occasions (number of daily snacking occasions), mean snack size (kcal/occasion), and mean snack energy density (kcal/g/occasion). Dietary weights were used to adjust for dietary nonresponse and day of the week [30]. Based on work done in a previous study by this team, initial models were evaluated with and without snacks containing trivial energy (eating occasions <5 kcal) [20]. Results were similar, therefore snack occasions with <5 kcal were excluded.
The Food Patterns Equivalents Database (FPED) was used to convert the What We Eat in America (WWEIA) dietary data to the USDA Food Pattern components [33]. Teaspoons of added sugar were converted to grams by multiplying by 4.2 and then grams were converted to calories by multiplying by 4; grams of saturated fat were converted to calories by multiplying by 9; sodium was reported in milligrams (mg).

Dietary Reporting Accuracy
The ratio of reported energy intake to estimated energy requirements (EI/EER) was used to assess dietary reporting accuracy. Estimated energy requirements were calculated using Dietary Reference Intake equations based on age, gender, weight, and physical activity level [34]. A "low active" level of physical activity (≥1.4 to <1.6) was assumed for all adolescents [35]. Following the methods in Murakami and Livingstone, the EI/EER ratio was used as a covariate in the analyses to account for EI reporting error (rather than removing implausible cases) so as not to bias sample selection [19].

Analyses
Descriptive statistics were generated and presented as means (standard errors) for continuous variables and percentages for categorical variables. Multiple linear regression models evaluated weight status as a predictor of each snacking parameter, adjusting for all other snacking parameters along with mean meal size (to account for non-snacking intake), survey cycle year and estimated energy reporting accuracy. Poisson regression using backward elimination was used to examine potential covariates. Child gender, age, race/ethnicity origin, RIP, survey cycle year, and HH age were retained and included as covariates in all models. Multiple linear regression models also examined the extent to which added sugar, saturated fats, sodium consumed from snacks varied by weight status. The primary outcome models examined total snacks including both foods and beverages. Supplementary models examined snacks from foods only and beverages only, separately. Although the central limit theorem posits that minor violations of normality of the dependent variables would not affect study conclusions, all analyses were repeated using 1000 bootstrapped replications to verify this; substantive conclusions were identical. All analyses were conducted in Stata (version 15.1; StataCorp, College Station, TX, USA).

Discussion
These nationally representative findings provide new evidence of the associations between snacking behaviors and weight status among US adolescents. The findings indicate that adolescents with OW and OB consume more snacks daily and more calories at each snacking occasion compared to adolescents who are classified as NW. Adolescents with OW and OB also consumed higher levels of added sugar, saturated fat, and sodium from snacks than adolescents with NW. By categorizing adolescents according to three weight classifications, normal weight, overweight and obesity, the findings suggest a potential dose response between snacking frequency and mean snack size and the association with weight status. Trends were consistent when examining associations between snacks from foods only and beverages only (Supplementary Materials, Table S1-S4, available online at www.mdpi.com/xxx/s1). These findings, while cross-sectional, provide population representative evidence of snacking associations with weight outcomes and suggest that approaches to obesity prevention that target snacking are relevant for adolescents.
Results are consistent with other studies that have used similar approaches to examine snacking frequency and weight status in children [19,20]. In a study by Murakami and Livingstone, snacking frequency was associated with higher risks of overweight and abdominal obesity in adolescents, after controlling for EI/EER. In a study examining associations between snacking frequency and weight status in younger children aged 1 to 5 years, positive associations were observed when snacking definitions took into account other eating occasions between meals (in addition to eating occasions identified as snacks) [20]. Based on this previous work, the current study expanded the definition of snacking to include self-identified snacking occasions plus other eating occasions between meals, and accounted for reporting bias in the analyses.
This study extends the current literature by examining snack size as an important dimension of snacking and the findings suggest that in addition to frequency, snack size seems to be important as it relates to obesity risk in adolescents. Taking into account the frequency and energy density of snacks consumed, adolescents at healthier weights are consuming smaller snacks. More work is needed to identify the optimal number and size of snacks for appropriate growth. Current dietary recommendations for adolescents do not clearly specify the number, size or type of snacks that should be consumed to promote optimal nutrient intake and manage weight.
Interestingly, the findings from this study do not provide evidence that adolescents with OW and OB are consuming more energy dense snacks compared to adolescents with NW, which is contrary to much of the existing literature on this topic [7,38]. After controlling for snack frequency and size, adolescents with OB consumed snacks that had a lower mean energy density per snack than children with NW. Energy density is defined as the number of calories per gram of food and can range from zero to nine, based on the composition of water and fat in the food (water has an energy density of zero kcal/g and fat has an energy density of 9 kcal/g) [39]. Energy density has been examined as a predictor of diet quality in children (e.g., lower energy density is associated with higher diet quality) and increasing intake of lower energy dense foods has been suggested as a strategy for weight management [40,41].
While energy density can act as a proxy for quality relevant to obesity, it does not fully capture nutritional quality or types of foods. In this study, mean snack energy density did not differ significantly between adolescents with NW and those with OW, and the difference between adolescents with NW and adolescents with OB was modest. On average, snack energy density in the total sample was 2.6 kcal/g/occasion, which is considered medium energy density and includes food items like breads, meats, cheeses, pretzels and popcorn [41]. Energy density in this range could reflect the consumption of these foods or of a combination of lower energy density foods (like fruits and vegetables) and nutrient-rich energy dense foods (like nuts and seeds). Since it is not possible to discern the types of foods from this study, further analyses are needed in order to characterize the types, sources and quality of snacks foods that adolescents are consuming.
This study contributes new evidence to the literature regarding snack intake and weight status in adolescents and has a number of strengths. First, three dimensions of snacking were examined including mean daily snacking occasions, mean snack size (kcal/occasion), and mean snack energy density (kcal/g/occasion). To our knowledge, this is the first study to include multiple snacking parameters in analyses examining the associations of snacking with weight status in adolescents. Multiple-pass 24-h dietary recalls were used to assess dietary intake and adolescents self-reported their intake [42]. While even this gold standard method is susceptible to bias, this study accounted for energy misreporting and evidence suggests that any misreporting most likely underestimated actual intake, especially in adolescents with higher weight status [43]. Findings from this study show that snacking varies by weight status, but also by gender and race/ethnicity. Additional studies are needed to examine context-specific factors that might differentially impact snacking behaviors these groups. For example, targeted marketing of snack foods might be influencing higher consumption among black adolescents and a deeper understanding could yield specific targets to improve dietary intake and address diet-related health disparities [44].
NHANES provides population-representative data in a large sample of US adolescents, but it is important to note limitations due to the cross-sectional design. Longitudinal research needs to be conducted to further elucidate the relationships found in this study. For example, snack frequency and snack size could be contributing to excess caloric intake and subsequent adiposity, but it is also possible that adolescents with overweight or obesity may have characteristics, such as impulsive eating, that make them more susceptible to snacking [45]. Further, while various approaches have been used to address measurement challenges, dietary measures are highly susceptible to reporting bias and the approach taken may not be sufficient to address those sources of measurement error. Another measurement limitation is the lack of a robust physical activity measure to account for energy expenditure. Exercise has been linked with compensatory snacking behavior, particularly unhealthy snacks, and this should be an area of future research, especially for growing children [46]. Lastly, the snacking measures from this study do not take into consideration the social, behavioral or environmental factors that may influence snack choices and these factors are important for developing salient targets for adolescents [47].

Conclusions
The nationally representative findings in this study provide evidence that US adolescents with OW and OB consume more snacks daily and more calories at each snacking occasion compared to adolescents with NW. Adolescents with OW and OB also consume greater amounts of added sugar, saturated fat and sodium from snacks than adolescents with NW, but consume snacks with a lower average energy density than NW adolescents. These findings suggest that snacking behaviors may be an important, multi-dimensional target for population-based approaches to reducing obesity risk among adolescents. Specifically, results suggest that the size of snacks consumed by adolescents as well as the frequency of daily snacking may have implications for weight status. Additional research is needed to determine the prospective association between snacking and weight status in adolescents. Future studies should also better characterize the types of foods that adolescents consume as snacks and how snacking behavior impacts overall diet quality. This information is critical for developing appropriate evidenced-based dietary recommendations for this age group and identifying effective obesity prevention and treatment targets.