SEARCH

SEARCH BY CITATION

Keywords:

  • snack foods;
  • soda;
  • weight;
  • television;
  • body composition

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgment
  8. References

Objective: The longitudinal relationship between the consumption of energy-dense snack (EDS) foods and relative weight change during adolescence is uncertain. Using data from the Massachusetts Institute of Technology Growth and Development Study, the current analysis was undertaken to examine the longitudinal relationship of EDS food intake with relative weight status and percentage body fat and to examine how EDS food consumption is related to television viewing.

Research Methods and Procedures: One hundred ninety-six nonobese premenarcheal girls 8 to 12 years old were enrolled between 1990 and 1993 and followed until 4 years after menarche. At each annual follow-up visit, data were collected on percentage body fat (%BF), BMI z score, and dietary intake. Categories of EDS foods considered were baked goods, ice cream, chips, sugar-sweetened soda, and candy.

Results: At study entry, girls had a mean ± SD BMI z score of −0.27 ± 0.89, consumed 2.3 ± 1.7 servings of EDS foods per day, and consumed 15.7 ± 8.1% of daily calories from EDS foods. Linear mixed effects modeling indicated no relationship between BMI z score or %BF and total EDS food consumption. Soda was the only EDS food that was significantly related to BMI z score over the 10-year study period, but it was not related to %BF. In addition, a significant, positive relationship was observed between EDS food consumption and television viewing.

Discussion: In this cohort of initially nonobese girls, overall EDS food consumption does not seem to influence weight status or fatness change over the adolescent period.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgment
  8. References

Obesity is the most significant nutritional problem facing children and adolescents in the United States. Obese children and adolescents experience a range of physical and psychological health consequences, including hypertension, decreased pulmonary function, sleep apnea, gallstone formation, insulin resistance, diabetes, low self-esteem, and eating disorders (1, 2). The dramatic rise in obesity prevalence among children of all ages, race/ethnicities, and socioeconomic backgrounds suggests that population-based preventive strategies are urgently needed to counter current trends and ensure that our nation's youths are not plagued by premature chronic health problems (3, 4).

At its most fundamental level, obesity is the result of an energy imbalance in which energy intake is greater than energy expenditure. Changes in diet, which affect energy intake, and changes in physical activity, which affect energy expenditure, alter energy balance and lead to weight change. Research on the activity and dietary behaviors of children indicates disturbances on both sides of the energy balance equation. The unhealthful eating patterns that lead to chronic disease in adulthood begin early in life, and diet-related chronic disease risk factors are prevalent among children in the United States (5). Several reports indicate that children are not eating the recommended amount of fruits and vegetables and that their food choices rarely follow the U.S. Department of Agriculture (USDA)1 food guide pyramid (6, 7, 8, 9). An analysis of data from four nationally representative USDA surveys showed a decrease in total energy intake and fat intake between 1965 and 1996 among adolescents 11 to 18 years of age. Over the same period, however, consumption of fruits, non-potato vegetables, and dairy foods decreased, whereas consumption of soft drinks increased (10). Mean energy intakes across the three National Health and Nutrition Examination Surveys (NHANES) either decreased slightly or remained stable for children 2 to 11 years of age but increased by ∼95 kcal/d for adolescents 12 to 19 years of age (11). Some authors have speculated that these findings indicate that increased energy intake is not likely to have been a major contributor to the increased prevalence of obesity observed over the same period (11).

National trends in snacking, assessed using data from the Continuing Survey of Food Intake of Individuals, indicate that the prevalence of snacking increased in all age groups between 1977 and 1996 (12). Furthermore, although the average size of a snack and the energy per snack has stayed relatively constant, the number of snacking occasions has increased significantly (12). An increase in the consumption of foods with added sweeteners is also of concern. National data indicate that the majority of added sweeteners in the diets of boys and girls 12 to 17 years of age comes from soda and fruit drinks (13) and that overall, male and female adolescents average an intake of 20% of total energy from added sweeteners (13). In an analysis of the dietary sources of nutrients among U.S. children 2 to 18 years of age, Subar et al. (14) found that cakes/cookies/quick breads/donuts appeared in the top 10 sources of energy, carbohydrate, protein, fat, saturated fat, fiber, and cholesterol. Soft drinks/soda were in the top 10 sources of energy and carbohydrates (14). Among adolescents, the major beverage contributor to energy was soft drinks, providing ∼8% of total energy (11).

On the energy expenditure side, many children and adolescents in the United States do not meet current national guidelines for regular physical activity (15, 16). In addition, leisure-time inactivity among youth is of concern. Television viewing is a major sedentary activity among children in the United States. Recent estimates indicate that children 2 to 18 years of age watch at least 2.5 hours of television each day and are exposed to a total of 6.5 hours of media per day from all sources (17, 18). Television viewing may promote increased energy intake from food consumed while watching television or from the effects of food advertising on food choices (19, 20). Studies have found an association between television viewing and increased energy intake and increased consumption of less healthful foods (21, 22).

Most cross-sectional studies of the relation between energy intake and obesity have shown that after adjusting for body weight or lean body mass, obese children do not eat more than their nonobese counterparts (23, 24, 25, 26, 27). Such studies are limited, however, by the differential under-reporting of energy intake that has been observed among obese compared with nonobese children (28, 29). Longitudinal data from a large cohort of adolescents showed that girls who reported higher energy intakes, less physical activity, and more television viewing had larger increases in BMI over 1 year (30).

The relationship between the type of food, rather than total energy intake, and obesity has also been explored. Troiano et al. (11) found that the contribution of soft drinks to energy intake was higher among overweight children. In a prospective analysis, Ludwig et al. (31) observed an association between the consumption of sugar-sweetened beverages and the odds of becoming obese over the follow-up period. In contrast, other studies do not support the notion that overweight adolescents eat more “junk food” than nonobese adolescents (32, 33).

Whereas data seem to indicate an increase in snacking among youth, the relationship between snack food consumption and body weight remains controversial. In particular, data on the relationship between the consumption of energy-dense snack (EDS) foods and changes in body weight over the adolescent period are lacking. In this study, we examined the EDS foods that we considered to be commonly consumed by adolescents: baked goods, ice cream, chips, sugar-sweetened soda, and candy. The objective of this analysis was to examine the relationship between EDS food consumption, weight status, and body fat in girls from pre-adolescence through adolescence, using annual data from a 10-year longitudinal study of growth and development in girls. The relationship between EDS food consumption and television viewing was also assessed.

Research Methods and Procedures

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgment
  8. References

Study Sample

The data for these analyses come from the Massachusetts Institute of Technology (MIT) Growth and Development Study, a prospective study designed to examine the relationship of energy expenditure to growth and development in girls from pre-adolescence to adolescence. Girls were recruited between the fall of 1990 and the spring of 1993. All fourth and fifth graders in Cambridge, MA, public schools were invited to participate; additional subjects were recruited from the MIT summer day camp and through contact with friends and siblings of subjects. At study entry (baseline), all girls were between 8 and 12 years old, premenarcheal, and nonobese based on a triceps skinfold thickness ≤85th percentile for age and sex according to NHANES I (34). This was a standard approach for the identification of overweight at the time the study was designed in the late 1980s.

All subjects were in good health as assessed by physical examination and medical histories. At study entry, the racial/ethnic composition of the cohort was 75% white, 14% black, and 11% other races. Either a study physician or a female co-investigator used Tanner's criteria of breast development to determine sexual maturation. On the anniversary of their baseline visit, participants returned for measurements every year until 4 years after menarche (study exit). If girls had not started menses at their annual visit, they were encouraged to telephone when they experienced their first menstrual period. Because most girls did not call, they were queried regarding menarche at each follow-up visit until menarche was reported. The study was approved by the Committee on the Use of Humans as Experimental Subjects at MIT and by the Human Investigations Review Committee of the New England Medical Center. The cohort consisted of 196 subjects at baseline.

Dietary Assessment

All subjects completed a Willett semiquantitative food frequency questionnaire (FFQ) at each annual follow-up visit. The version of the FFQ available when the study began in the early 1990s was used throughout the study. The questionnaire was specially designed for children based on a validated semiquantitative FFQ for adults (35). Similar to the adult version, the questionnaire was designed to be self-administered. However, subjects were given verbal and/or written instructions on how to properly complete the forms. The 116-item FFQ was based on recall of diet in the past year. EDS foods considered in this analysis include cookies, pies, cakes, brownies, chocolate candy, nonchocolate candy, ice cream, ice cream sundaes, milkshakes, sherbet, potato chips, corn chips, and soda. Serving sizes were of natural units or typical servings sizes. When completing the FFQ, subjects indicated how often, on average, they had consumed the amount of each food item in the past year. The nine response categories available ranged from “never or <1 per month” to “6 or more per day.” The individual EDS foods considered in this analysis were divided into five food categories: 1) baked goods (cookies, cakes, pies, and brownies); 2) ice cream (ice cream, ice cream sundaes, sherbet, milkshakes); 3) chips (potato chips and corn chips); 4) candy (chocolate and nonchocolate candy); and 5) soda (only sugar-sweetened). The ice cream category was included because of its high calorie content.

The food composition database used to calculate levels of intake for calories and nutrients was based on publications from the USDA, laboratories, and manufacturers. Calories and nutrient intakes were calculated by multiplying the frequency of consumption by the nutrient composition for the portion size of each specific food listed. Calories and nutrients were summed across all foods to obtain total levels of all calories and nutrients for each subject. Servings of specific EDS foods were converted into daily servings, and total daily EDS food servings were calculated by summing across all EDS foods. The percentage of daily kilocalories from EDS foods was calculated by adding the calories from each EDS food and dividing the sum by total daily kilocalories. Dietary variables considered as potential covariates (servings of fruits and vegetables and percentage of calories from protein, fat, and carbohydrate) were calculated from the FFQ as well.

Other analyses conducted in this cohort have indicated that the FFQ provides a reasonable estimate of EDS food intake. We found that EDS food consumption estimated from the FFQ correlated well with EDS food consumption reported on 7-day diet records at baseline and study exit. (Spearman's R at study exit for snack food servings is 0.33; correlation is between FFQ estimate and diet record estimate, unpublished observations.)

Anthropometry

Height and body weight were measured in the morning. Height was measured to 0.1 cm with a wall-mounted stadiometer. Weight was measured with subjects in a hospital gown using a Seca scale (Hanover, MD) accurate to 0.1 kg. BMI was calculated as weight in kilograms per height in meters squared. BMI z score was calculated using the Centers for Disease Control and Prevention (CDC) modified growth reference standards (36). Bioelectrical impedance analysis (BIA) was used to measure resistance (R) and reactance after an overnight fast or a 2-hour postprandial (BIA 101; RJL Systems, Clinton Township, MI). The accuracy of the machine was checked before the measurement with a 500-ω resistor supplied by the manufacturer. Measurements were taken with the subject supine, and electrodes were placed on the dorsal surface of the right foot and ankle and right wrist and hand. A current was applied at a frequency of 50 kHz. Percentage body fat (%BF) was estimated using prediction equations developed in this cohort, with measures of total body water by isotopic dilution of H218O as the criterion method. Separate equations were used depending on the menarcheal status of the participant. We found that %BF estimated from our equation closely approximated %BF estimated by H218O in our cohort (37).

Physical Activity Measures

Subjects completed a questionnaire at each annual follow-up visit, identifying usual patterns of physical activity. Subjects were presented with two 24-hour timetables and asked to recall, on an hourly basis, their participation in five types of activities during each time block: sleeping or lying down, sitting, standing, walking, and vigorous activity (exercising, playing, or being involved in sports). In addition, subjects completed a similar grid on which they reported television viewing time on an hourly basis (including time spent watching videos or playing video games). Separate time blocks were completed for school and weekend days. The average daily time spent in each activity was computed as a weighted average of the school day and weekend day reports. Average daily times spent walking and in vigorous activity were combined and weighted by their intensity (using a MET value) to create an activity index, which was calculated as 2.5 × walk + 5.5 × play. Average daily times spent sleeping or lying, sitting, and standing were combined to create a variable representing total inactivity time (hours per day). Information on the reliability of this physical activity assessment protocol has been published elsewhere (38). In this cohort, the correlation between baseline nonresting energy expenditure and baseline physical activity index was 0.29.

Variables and Data Exclusions

The main exposure of interest was EDS food consumption, expressed as total servings per day, percentage of daily kilocalories from all EDS foods, and percentage of daily calories from each of the five categories of EDS foods considered in this analysis. When necessary, exposure variables were log-transformed to better approximate normality. In our analysis of the percentage of daily calories from specific EDS food categories, data were categorized into quartiles because the data exhibited extreme non-normality.

In analyses to investigate the relationship between EDS food consumption and television viewing, EDS food consumption (servings per day and percent of daily calories from EDS foods) was treated as the outcome, and log-transformed hours of television viewed per day were used as the exposure.

Dietary exclusion criteria were used to omit annual visits when subjects left more than 12 items blank on the FFQ or when daily energy intake was <500 or >5000 kcals as calculated from the FFQ. In addition, subjects with fewer than three annual visits were excluded. Therefore, these analyses include data from 178 (91%) of the subjects, representing 1198 data points, with an average of 7.7 measurements per subject. (Not all subjects had data at baseline and/or exit.)

Statistical Analysis

Paired t tests and Wilcoxon signed-rank tests (for variables whose differences were non-normal) were used in the simple comparison of changes between baseline and exit. Generalized additive modeling (GAM) was used to visualize the relationship between the BMI z score or %BF and EDS food consumption (39). GAM models were run separately for outcome/predictor pairs. The predicted value from the GAM and its 95% confidence interval was plotted on a scale to cover 95% of the distribution for a given outcome. Although this technique ignores the correlation structure in repeated measurements, these plots allowed us to assess the appropriateness of using linear mixed effects (LME) modeling and to visualize the general pattern of the relationship.

LME modeling was used to evaluate the longitudinal relationship between relative body weight or body fatness and EDS food consumption. Because we had two outcomes (BMI z score and %BF) and seven exposure variables (daily servings of EDS foods, percent daily calories from EDS foods, and percentage of daily calories from each of the five categories of EDS foods), 14 separate LME models were evaluated. The applied mixed effects model consists of two parts: fixed and random effects. Fixed effects describe a population intercept and population slopes for a set of considered covariates, which include exposures and confounders. Random effects describe individual variability in the outcome and changes over time. By considering individual random slopes and intercepts, this model allows us to examine the influence of covariates on the change in outcome over time. The LME model also accounts for the correlation between repeated measurements on the same subject and the different numbers of measurements per subject.

To control for possible confounders in the relationship between either BMI z score or %BF and EDS food consumption, longitudinal models were evaluated to determine which potential covariates were significant predictors of EDS food consumption and either BMI z score or %BF. The following variables were considered: physical activity index, inactivity time, parental overweight (defined as at least one parent with a BMI > 25 kg/m2), race/ethnicity (coded as two dummy variables for black and “other,” with white as the reference category), daily servings of fruits and vegetables, percentage of daily calories from protein, percentage of daily calories from carbohydrates, and percentage of daily calories from fat. For models with BMI z score as the outcome variable, age was expressed as chronological age, and age at menarche was included as a fixed covariate. For models with %BF as the outcome variable, age was expressed relative to age at menarche (40). With the exception of age and parental overweight, we included as covariates only those variables that were significant predictors of both the exposure and the outcome. This approach allowed us to streamline the inclusion of covariates and made the model more parsimonious, thereby avoiding the risk of overparamaterizing the model. Parental overweight was included in all models because of its strong longitudinal relationship with BMI z score and %BF. In our analysis of the percentage of daily calories from specific EDS food categories, quartiles of percentage calories from each EDS food category were based on consumption levels at baseline. The boundaries for these quartile values were applied to all of the measurements. The only exception to this approach was in two models with %BF and the percentage of calories from baked goods and candy, where quartiles were replaced by tertiles so that there were sufficient counts in all categories for model convergence. In the longitudinal models, quartiles and tertiles were treated as dummy variables, with the lowest category as the reference. Data were analyzed using SAS (Version 8.0; SAS Institute, Cary, NC) and S-PLUS (Version 4.5; MathSoft Inc., Seattle, WA) software. α was set at 0.05 for all analyses.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgment
  8. References

Characteristics of Study Sample

Characteristics of the cohort at baseline and exit are shown in Table 1. Sixty-five percent of girls were Tanner stage 1, and 35% were Tanner stage 2 or 3 at baseline. The cohort was predominantly white, with an average age of ∼10 years at baseline and 17 years at exit. The mean ± SD BMI z score was −0.27 ± 0.89 at baseline, reflecting the study entry criteria. Significant increases in BMI z score (p < 0.01) and %BF (p < 0.01) were observed between study entry and study exit. Subjects consumed an average of 2.3 servings of EDS foods at baseline and an average of 2.0 servings of EDS foods at exit (t = 2.0, p = 0.04). Subjects consumed nearly 16% of their daily calories from EDS foods at both baseline and exit (15.7% and 15.6%, respectively).

Table 1. . Characteristics of the cohort at study entry and 4 years after menarche (study exit)
 Study entry (n = 166)Study exit (n = 141)Paired comparison (n = 132)
CharacteristicMean (SD)Median*Mean (SD)Median*p
  • *

    Medians presented for variables whose distributions were non-normal.

  • Paired comparison based on nonparametric Wilcoxon signed-rank test.

Age (years)10.0 (0.93)9.916.9 (1.0)17.0<0.001
BMI16.6 (1.8)16.521.4 (2.5)21.1<0.001
BMI z-score−0.27 (0.89)−0.250.02 (0.79)0.07<0.001
Percent body fat by BIA23.4 (4.7)23.127.6 (3.8)27.8<0.001
TV (hours/day)3.5 (2.5)3.01.8 (1.5)1.4<0.001
Activity time (hours/day)4.1 (1.6)4.13.7 (1.8)3.4<0.001
Inactivity time (hours/day)19.9 (1.6)19.920.3 (1.8)20.6<0.001
Daily kilocalories2021 (669)19331723 (655)1640<0.001
Percent calories from fat30.1 (5.3)30.527.9 (6.2)27.6<0.001
Percent calories from protein15.6 (3.0)15.616.3 (3.4)16.40.04
Percent calories from carbohydrates56.3 (7.1)55.757.6 (8.2)57.70.08
EDS foods (servings/day)2.3 (1.7)1.862.0 (1.4)1.70.04
Percent daily kcal from EDS foods15.7 (8.1)14.515.6 (8.1)14.20.78
Percent daily kcal from baked goods3.6 (2.3)3.43.4 (2.6)2.90.78
Percent daily kcal from ice cream3.6 (3.8)3.03.1 (3.2)2.50.14
Percent daily kcal from chips2.2 (2.3)1.51.4 (1.6)1.0<0.001
Percent daily kcal from soda2.7 (3.9)1.64.0 (4.6)2.50.002
Percent daily kcal from candy3.5 (3.7)2.63.8 (3.4)3.10.29

Longitudinal Analyses

Changes in EDS Food Consumption, Television Viewing, and Physical Activity with Age

There was a borderline significant decrease in daily servings of EDS foods with increasing age (p = 0.05). Based on our model, between ages 10 and 16 years, daily servings of EDS foods would decrease by ∼0.20 servings. When EDS food consumption was expressed as a percentage of daily kilocalories from EDS foods, no significant relationship with age was observed (p > 0.05). Physical activity (assessed by the physical activity index) and television viewing decreased significantly (p < 0.001), whereas inactivity (hours per day) increased significantly (p < 0.001).

Relationship between EDS Food Consumption and BMI z score

We found no statistically significant relationship between total EDS food consumption, expressed as servings per day or as percentage of daily calories, and BMI z score (Table 2). Within the specific groups of EDS foods, we found no significant relationship between the percentage of calories from ice cream, baked goods, candy, or chips (expressed as quartiles) and BMI z score (Table 3). We did observe a significant relationship between the percentage of calories from soda and BMI z score, which remained significant after adjusting for covariates (Table 3). Subjects in the third and fourth quartiles of percentage calories from soda had BMI z scores that were ∼0.17 units higher on average than subjects in the first quartile. When the data were stratified by menarcheal status (pre- vs. postmenarche), the relationship between BMI z score and soda consumption was significant during only the postmenarcheal period.

Table 2. . Results from linear mixed models predicting BMI z score from EDS food consumption
 Estimatep
  • *

    Adjusted for age at menarche, daily servings of fruits and vegetables, and parental overweight.

Daily servings  
 Model 1  
  Intercept−0.74<0.001
  Age0.05<0.001
  Log daily servings of snack foods−0.00240.94
 Model 2  
  Intercept2.92<0.001
  Age0.05<0.001
  Log daily servings of snack foods0.0350.33
Percent of daily kcal  
 Model 1  
  Intercept−0.79<0.001
  Age0.05<0.001
  Percent daily kcal from snacks0.00250.19
 Model 2*  
  Intercept2.82<0.001
  Age0.05<0.001
  Percent daily kcal from snacks0.00360.11
Table 3. . Results from linear mixed models predicting BMI z score from quartiles of percentage of calories from various snack foods
ExposureEstimatepp for trend
  • *

    Adjusted for age at menarche, parental overweight, and servings of fruits and vegetables.

Percentage of calories from soda*   
 Intercept3.12<0.001 
 Age0.05<0.001 
 First quartile (<0.74)Referent <0.001
 Second quartile (0.75 to 1.4)0.0890.030 
 Third quartile (1.5 to 3.1)0.172<0.001 
 Fourth quartile (≥3.2)0.1780.001 
Percentage of calories from candy*   
 Intercept3.06<0.001 
 Age0.05<0.001 
 First quartile (<1.5)Referent 0.088
 Second quartile (1.5 to 2.4)0.0210.52 
 Third quartile (2.5 to 3.9)0.0050.90 
 Fourth quartile (≥4.0)0.0820.066 
Percentage of calories from chips*   
 Intercept2.78<0.001 
 Age0.05<0.001 
 First quartile (<1.0)Referent 0.24
 Second quartile (1.0 to 1.4)0.0320.39 
 Third quartile (1.5 to 2.4)0.0300.39 
 Fourth quartile (≥2.5)0.0470.26 
Percentage of calories from baked goods*   
 Intercept2.97<0.001 
 Age0.05<0.001 
 First quartile (<2.0)Referent 0.33
 Second quartile (2.0 to 3.4)−0.0290.31 
 Third quartile (3.5 to 4.4)−0.0300.45 
 Fourth quartile (≥4.5)−0.0270.42 
Percentage of calories from ice cream*   
 Intercept2.72<0.001 
 Age0.05<0.001 
 First quartile (<1.0)Referent 0.85
 Second quartile (1.0 to 2.9)0.000540.98 
 Third quartile (3.0 to 3.9)0.01240.14 
 Fourth quartile (≥4.5)−0.00920.82 
Relationship between EDS Food Consumption and %BF

We found no statistically significant relationship of total EDS food consumption, expressed as servings per day or as a percentage of daily calories, with %BF (Table 4). In addition, we found no significant relationship between the percentage of calories from soda, ice creams, candy, baked goods, or potato chips (expressed as quartiles or tertiles) and %BF (Table 5).

Table 4. . Results from linear mixed models predicting % BF from EDS food consumption
 Estimatep
  • *

    Age expressed as age relative to menarche.

  • Adjusted for percentage of calories from protein and parental overweight.

Daily servings  
 Model 1  
  Intercept25.1<0.001
  Age*0.58<0.001
  Log daily servings of snack foods−0.130.63
 Model 2  
  Intercept22.2<0.001
  Age*0.57<0.001
  Log daily servings of snack foods0.200.49
Percent of daily kcal  
 Model 1  
  Intercept24.9<0.001
  Age*0.58<0.001
  Percent daily kcal from snacks0.00640.66
 Model 2  
  Intercept21.8<0.001
  Age*0.57<0.001
  Percent daily kcal from snacks0.0280.13
Table 5. . Results from linear mixed models predicting %BF from quantiles of percentage calories from various snack foods
ExposureEstimatepp for trend
  • *

    Adjusted for parental overweight and percentage of calories from protein.

Percentage of calories from soda*   
 Intercept22.3<0.001 
 Age0.58<0.001 
 First quartile (<0.74)Referent 0.23
 Second quartile (0 to 1.4)0.150.57 
 Third quartile (1.5 to 3.1)0.410.18 
 Fourth quartile (≥3.2)0.310.35 
Percentage of calories from candy*   
 Intercept22.6<0.001 
 Age0.60  
 First tertile (<2.0)Referent 0.35
 Second tertile (2.0 to 3.4)−0.0510.52 
 Third tertile (≥3.5)0.0660.96 
Percentage of calories from chips*   
 Intercept22.5<0.001 
 Age0.57<0.001 
 First quartile (<1.0)Referent 0.63
 Second quartile (1.0 to 1.4)−0.0540.86 
 Third quartile (1.5 to 2.4)−0.1010.66 
 Fourth quartile (≥2.5)−0.1610.56 
Percentage of calories from baked goods*   
 Intercept24.1<0.001 
 Age0.59<0.001 
 First tertile (<2.5)Referent 0.23
 Second tertile (2.5 to 3.9)−0.1030.57 
 Third tertile (≥4.0)−0.2210.23 
Percentage of calories from ice cream*   
 Intercept22.6<0.001 
 Age0.60<0.001 
 First quartile (<1.0)Referent 0.89
 Second quartile (1.0 to 2.9)−0.00870.37 
 Third quartile (3.0 to 3.9)0.400.46 
 Fourth quartile (≥4.5)0.190.82 
Relationship between EDS Food Consumption and Activity, Inactivity, and Television Viewing

There was no relationship between EDS food consumption, expressed as servings per day or percentage calories, and physical activity or sedentary behavior. We did observe a significant relationship between hours of television viewed per day and EDS food consumption, expressed as either daily servings or as a percentage of daily calories from EDS foods (Table 6). The GAM plots depicting the relationship between EDS food consumption and television viewing are shown in Figure 1.

Table 6. . Results from linear mixed models predicting snack consumption from television viewing
 Estimatep
  • *

    Adjusted for race/ethnicity, percentage of calories from protein, percentage of calories from fat, and parental overweight.

Log daily snack servings*  
 Intercept1.08<0.001
 Age0.0040.44
 Log daily television hours0.126<0.001
Percent daily kcal from snacks*  
 Intercept13.6<0.001
 Age0.33<0.001
 Log daily television hours1.87<0.001
image

Figure 1. Relationship between EDS food consumption and television viewing. HC, high calorie.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgment
  8. References

Although a myriad of factors have likely contributed to the current epidemic of childhood obesity, successful prevention strategies must focus on the etiologic components that are most amenable to change. The role of specific types of foods in the promotion of overweight and obesity among children represents one potential intervention target and is the subject of much debate. In this study, we examined intakes of EDS foods commonly consumed by adolescents: chips, soda, candy, baked goods, and ice cream. Using longitudinal data collected annually over a 10-year period, we observed no significant relationship between total EDS food consumption and BMI z score or %BF over the adolescent period. Although total EDS food consumption was not related to body weight or fatness, we observed a significant relationship between soda consumption and BMI z score, but not %BF, over the 10-year study period. We also observed a significant relationship between EDS food consumption and television viewing.

Whereas several studies have examined differences in total energy intake among obese and nonobese children, few have examined the consumption of specific foods in relation to weight status among youth. In cross-sectional analyses, energy intake has been positively associated with consumption of nondiet soft drinks (9). In a cross-sectional analysis, Bandini et al. (30) examined the intake of EDS foods among obese and nonobese adolescents. We found that the consumption of EDS foods (expressed as percentage of total calories) was similar between the obese and nonobese groups after adjustment for degree of under-reporting (32). Gibson (33) observed a significant inverse association between the percentage of calories from biscuits/cakes, sugar, and preserves and BMI, but no significant association between the percentage of calories from soft drinks and BMI was found. Cross-sectional data from the Third NHANES showed that the contribution of soft drinks to energy intake was higher among overweight children (11). Another analysis of data from NHANES III found no association between nondiet soft drink consumption and BMI in adolescents 12 to 16 years of age (41). In a 19-month prospective study, Ludwig et al. (31) observed that both baseline sugar-sweetened beverage consumption and change in consumption independently predicted change in BMI over the study period. Their analysis, however, used change in BMI rather than BMI z score as a measure of weight change.

Although soda has not been tested exclusively in an experimental setting, the effect of sucrose on body weight in adults was recently explored in an intervention study in which 41 overweight subjects were randomized to consume daily supplements of either sucrose or artificial sweeteners for 10 weeks (∼80% by weight of the supplements were beverages) (42). After 10 weeks, the sucrose group had significant increases in body weight and fat mass compared with the artificial sweetener group. The mechanism by which sugar-containing drinks may influence weight status is uncertain. One possibility is that compensation for energy consumed in liquid form is less complete than for energy consumed in a solid form (31, 42).

In our data, it is not clear why the relationship we observed between soda consumption and BMI z score was not seen when %BF was the outcome variable. One possibility is that more active girls drink more soda, and their relatively higher BMI z score is a reflection of increased muscle, rather than fat, mass. In our cohort, however, physical activity was not significantly related to soda consumption. Alternatively, early maturing girls would have a higher BMI z score than their later maturing peers. If soda consumption was a marker for adolescent behavior, soda consumption would be higher in early maturing girls and would seem to be related to BMI z score. In our models, however, adjusting for age relative to menarche rather than chronological age did not alter the results.

Although we observed no relationship between EDS food consumption and either physical activity or sedentary behavior, the significant relationship of television viewing to EDS food consumption is noteworthy. The absence of a significant relationship between snacking and overall sedentary behaviors suggests that there is something unique about television viewing and its contribution to EDS food consumption. Television viewing has been associated with increased energy intake and poorer diet quality. Crespo et al. (21) found that energy intake tended to increase with increased television viewing among children 8 to 16 years of age. Another recent study found that children from families who watch television during two or more meals a day consume more red meat, pizza, soda, and salty snacks (22). Increased food consumption during television viewing may be due, in part, to exposure to food advertising during television programming. Television is the single largest media source of messages about food (22). Several studies have found that children's request for and consumption of advertised foods and parental willingness to purchase them are positively related to the number of hours of television viewed by children (43, 44, 45, 46, 47, 48).

We believe our approach has several strengths. Our analyses are based on a large number of annual measurements taken repeatedly over the adolescent period. The analytic approach selected allows us to capitalize on the richness of these data by characterizing individual variation relative to the population mean while taking into account the correlation between repeated measurements on the same subject and different numbers of measurements per subject. Furthermore, additional analyses conducted in this cohort indicate that both the FFQ and BIA provide good estimates of EDS food intake and %BF, respectively.

Our analysis also has some important limitations. Although the foods we considered are typical snack foods, we could not classify them as snack foods because we lacked information on whether they were eaten with or between meals. As such, we cannot address the issue of “snacking” as it relates to patterns of meal consumption. Second, the FFQ, like all dietary assessments, are subject to considerable error. Nonetheless, although not directly comparable because of differences in age categorization and in dietary methodology, daily energy intakes estimated by the FFQ in our cohort compared reasonably well with the 1989 kcal reported for 12- to 15-year-old girls based on 24-hour recall in NHANES III. Third, differential reporting of food intake is a concern in any study examining the relationship between self-reported food intake and body weight. Indeed, many studies that have observed either no difference in energy intake between the obese and nonobese or lower energy intakes among the obese have been criticized for this reason. Perks et al. (49) compared energy intake estimated by FFQ with total energy expenditure measured by doubly labeled water in children and adolescents. They found that energy intake reported on the FFQ and energy expenditure by doubly labeled water were similar. However, discrepancy in energy intake was related to body weight and BF (49). Differential reporting of specific foods is more difficult to study. Krebs-Smith et al. (50) found that foods high in added sugars are selectively under-reported. Soda consumption in this cohort was lower than soda consumption reported in nationally representative surveys. In NHANES III, soda consumption (as a percentage of energy intake) was 4.1% for all children 2 to 5 years of age and 7.9% for 12- to 19-year-old girls compared with 2.7% in our cohort at baseline and 4.1% at study exit (11). Although adolescence is considered a critical period for the development of obesity in girls, our exclusion at baseline of any girls who were already overweight precludes our ability to examine the role of EDS foods on weight or fatness in girls who are already overweight in pre-adolescence. Last, all dietary methodologies are subject to measurement error; however, a validation conducted in this cohort indicates that the FFQ does a reasonably good job of estimating EDS food consumption (compared with consumption from a 7-day food diary; S. M. Phillips, unpublished observations).

In summary, weight gain over the adolescent period was not related to increased overall consumption of EDS foods in this cohort of initially nonobese girls. A possible exception to this overall finding is soda consumption, where we saw a significant positive longitudinal relationship with BMI z score. We did not see a similar effect of soda consumption on body fatness changes over adolescence. Although the effect size for the estimated impact of soda consumption on BMI z score is not large, it could be important on a population basis, given the rising trend in soda consumption,. Furthermore, aside from any potential impact on weight status, high consumption of EDS foods may be of concern because of their low nutrient density. These data also reinforce the notion that limiting TV time may modify EDS food consumption. Further longitudinal studies of children of other age groups, across the full range of weight status, and of boys are needed improve our understanding of the role of specific foods in weight and fatness changes during childhood.

Acknowledgment

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgment
  8. References

This study was conducted at the General Clinical Research Center at Massachusetts Institute of Technology and supported by National Institutes of Health Grants MOI-RR-00088, DK-HD50537, and 5P30 DK46200 and by Mars, Inc. This study was initiated before W.H.D's employment at the CDC. The conclusions contained herein do not reflect CDC policy. We thank Zoom Compton, Tara Mardigan, and the staff at the Clinical Research Center for assistance with the study and the girls who enrolled for their participation and commitment.

Footnotes
  • 1

    Nonstandard abbreviations: USDA, U.S. Department of Agriculture; NHANES, National Health and Nutrition Examination Study; EDS, energy dense snack; MIT, Massachusetts Institute of Technology; FFQ, food frequency questionnaire; CDC, Center for Disease Control and Prevention; BIA, bioelectrical impedance analysis; %BF, percentage body fat; GAM, generalized additive modeling; LME, linear mixed effects.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Research Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgment
  8. References
  • 1
    French, S. A., Story, M., Perry, CL. (1995) Self-esteem and obesity in children and adolescents: a literature review. Obes Res. 3: 479490.
  • 2
    Must, A., Strauss, RS. (1999) Risks and consequences of childhood and adolescent obesity. Int J Obes Relat Metab Disord. 23: S2S11.
  • 3
    Schmitz, M. K., Jeffery, RW. (2000) Public health interventions for the prevention and treatment of obesity. Med Clin North Am. 84: 491511.
  • 4
    Ogden, C. L., Flegal, K. M., Carroll, M. D., Johnson, CL. (2002) Prevalence and trends in overweight among US children and adolescents, 1999–2000. JAMA. 288: 17281732.
  • 5
    Center for Disease Control and Prevention. (1996) Guidelines for school health programs to promote lifelong healthy eating. MMWR Morb Mortal Wkly Rep. 45: 133.
  • 6
    Devaney, B. L., Gordon, A. R., Burghardt, JA. (1995) Dietary intakes of students. Am J Clin Nutr. 61: 205S212S.
  • 7
    U. S. Department of Agriculture, Agricultural Research Service. (Accessed February 24, 2004) Data Tables: Results from USDA's 1994–96 Continuing Survey of Food Intakes by Individuals. http:www.barc.usda.govbhnrcfoodsurveypdf3yr_py.pdf.
  • 8
    Siega-Riz, A., Popkin, B. M., Carson, T. (1998) Trends in breakfast consumption for children in the United States from 1965 to 1991. Am J Clin Nutr. 67: 748S756S.
  • 9
    Harnack, L., Stang, J., Story, M. (1999) Soft drink consumption among U. S. children and adolescents: nutritional consequences. J Am Diet Assoc. 99: 436441.
  • 10
    Cavadini, C., Siega-Riz, A. M., Popkin, BM. (2000) U.S. adolescent food intake trends from 1965 to 1996. West J Med. 173: 378385.
  • 11
    Troiano, R. P., Briefel, R. R., Carroll, M. D., Bialostosky, K. (2000) Energy and fat intakes of children and adolescents in the United States: data from the National Health and Nutrition Examination Surveys. Am J Clin Nutr. 72: 1343S1353S.
  • 12
    Jahns, L., Siega-Riz, A. M., Popkin, BM. (2001) The increasing prevalence of snacking among US children from 1977 to 1996. J Pediatr. 138: 493498.
  • 13
    Guthrie, J. F., Morton, JF. (2000) Food sources of added sweeteners in the diets of Americans. J Am Diet Assoc. 100: 4351.
  • 14
    Subar, A. F., Krebs-Smith, S. M., Cook, A., Kahle, LL. (1998) Dietary sources of nutrients among U. S. children, 1989–1991. Pediatrics. 102: 913923.
  • 15
    Caspersen, C. J., Pereira, M. A., Curran, KM. (2000) Changes in physical activity patterns in the United States, by sex and cross-sectional age. Med Sci Sports Exerc. 32: 16011609.
  • 16
    Schlicker, S., Borra, S., Regan, C. (1994) The weight and fitness status of United States children. Nutr Rev. 52: 1117.
  • 17
    Woodard, E. H., Gridina, N. (2000) Media in the Home: 2000. The Annenberg Public Policy Center of the University of Pennsylvania: Philadelphia, PA. 1926.
  • 18
    Roberts, D. F., Foehr, U. G., Rideout, V. J., Brodie, M. (1999) Kids and Media at the New Millennium. A Henry J. Kaiser Family Foundation Report. Henry J. Kaiser Foundation Menlo: Park, CA. 2728.
  • 19
    Dietz, W. H., Gortmaker, SL. (2001) Preventing obesity in children and adolescents. Annu Rev Public Health. 22: 237253.
  • 20
    Robinson, TN. (2001) Television viewing and childhood obesity. Pediatr Clin North Am. 48: 10171025.
  • 21
    Crespo, C. J., Smit, E., Troiano, R. P., Bartlett, S. J., Macera, C. A., Andersen, RE. (2001) Television watching, energy intake, and obesity in US children: results from the third National Health and Nutrition Examination Survey, 1988–1994. Arch Pediatr Adolesc Med. 155: 360365.
  • 22
    Coon, K. A., Goldberg, J., Rogers, B. L., Tucker, KL. (2001) Relationships between use of television during meals and children's food consumption patterns. Pediatrics. 107: E7
  • 23
    Rocandio, A. M., Ansotegui, L., Arroyo, M. (2001) Comparison of dietary intake among overweight and non-overweight school children. Int J Obes Relat Metab Disord. 25: 16511655.
  • 24
    McGloin, A. F., Livingstone, M. B. E., Greene, L. C., et al (2002) Energy and fat intake in obese and lean children at varying risk of obesity. Int J Obes Relat Metab Disord. 26: 200207.
  • 25
    Garaulet, M., Martinez, A., Victoria, F., Perez-Llamas, F., Ortega, R. M., Zamora, S. (2000) Differences in dietary intake and activity level between normal-weight and overweight or obese adolescents. J Pediatr Gastroenterol Nutr. 30: 253258.
  • 26
    Maffeis, C., Talamini, G., Tato, L. (1998) Influence of diet, physical activity, and parents’ obesity on children's adiposity: a four-year longitudinal study. Int J Obes Relat Metab Disord. 22: 758764.
  • 27
    Gillis, L. J., Kennedy, L. C., Gillis, A. M., Bar-Or, O. (2002) Relationship between juvenile obesity, dietary energy and fat intake and physical activity. Int J Obes Relat Metab Disord. 26: 458463.
  • 28
    Bandini, L. G., Schoeller, D. A., Cyr, H., Dietz, WH. (1990) Comparison of energy intake and energy expenditure in obese and nonobese adolescents. Am J Clin Nutr. 52: 421425.
  • 29
    Livingstone, M. B. E., Prentice, A. M., Coward, WA. (1992) Validation of estimates of energy intake by weighed dietary record and diet history in children and adolescents. Am J Clin Nutr. 56: 2935.
  • 30
    Berkey, C. S., Rockett, H. R., Field, A. E., et al (2000) Activity, dietary intake, and weight changes in a longitudinal study of preadolescent boys and girls. Pediatrics. 105: E56
  • 31
    Ludwig, D. S., Peterson, K. E., Gortmaker, SL. (2001) Relation between consumption of sugar-sweetened drinks and childhood obesity: a prospective, observational analysis. Lancet. 357: 505508.
  • 32
    Bandini, L. G., Vu, D., Must, A., Cyr, H., Goldberg, A., Dietz, WH. (1999) Comparison of high-calorie, low-nutrient-dense food consumption among obese and non-obese adolescents. Obes Res. 7: 438443.
  • 33
    Gibson, S. (1998) Hypothesis: parents may selectively restrict sugar-containing foods for pre-school children with a high BMI. Int J Food Sci Nutr. 49: 6571.
  • 34
    Must, A., Dallal, G. E., Dietz, W. (1991) Reference data for obesity: 85th and 95th percentiles of body mass index (wt/ht2) and triceps skinfold thickness. Am J Clin Nutr. 53: 839846.
  • 35
    Willett, W. C., Sampson, L., Stampfer, M. J., et al (1985) Reproducibility and validity of a semiquantitative food frequency questionnaire. Am J Epidemiol. 122: 5165.
  • 36
    Ogden, C. L., Kuczmarski, R. J., Flegal, K. M., et al (2002) Centers for Disease Control and Prevention 2000 growth charts for the United States: improvements to the 1977 National Center for Health Statistics version. Pediatrics. 109: 4560.
  • 37
    Phillips, S. M., Bandini, L. G., Compton, D. V., Naumova, E. N., Must, A. (2003) A longitudinal comparison of body composition by total body water and bioelectrical impedance in adolescent girls. J Nutr. 133: 14191425.
  • 38
    Ching, P., Dietz, WH. (1995) Reliability and validity of activity measures in preadolescent girls. Pediatr Exerc Sci. 7: 389399.
  • 39
    Hastie, T. J., Tibshirani, RJ. (1990) Generalized Additive Models. Chapman & Hall: New York, NY.
  • 40
    Naumova, E. N., Must, A., Laird, NM. (2001) Tutorial in biostatistics: evaluating the impact of ‘critical Period’ in longitudinal studies of growth using piecewise mixed effects models. Int J Epidemiol. 30: 13321341.
  • 41
    Smith, P. A., Forshee, R. A., Storey, ML. (2001) Relationship between soft drink consumption and BMI among teens. FASEB J. 15: A1001 (abstr).
  • 42
    Raben, A., Vasilaras, T. H., Moller, C., Astrup, A. (2002) Sucrose compared with artificial sweeteners: different effects on ad libitum food intake and body weight after 10 wk of supplementation in overweight subjects. Am J Clin Nutr. 76: 721729.
  • 43
    Young, B., Hetherington, M. (1996) The literature on advertising and children's food choice. Nutr Food Sci. 5: 1518.
  • 44
    Taras, H. L., Sallis, J. F., Patterson, T. L., Nader, P. R., Nelson, JA. (1989) Television's influence on children's diet and physical activity. Dev Behav Pediatr. 10: 176180.
  • 45
    Gorn, JGE. (1982) Behavioral evidence of the effects of televised food messages on children. J Consumer Res. 9: 200205.
  • 46
    Galst, JPW. (1980) Television commercials and pronutritional public service announcements as determinants of young children's snack choices. Child Dev. 51: 935938.
  • 47
    Goldberg, M., Gorn, G., Gibson, W. (1978) TV messages for snack and breakfast foods: do they influence children's preferences? J Consumer Res. 5: 125130.
  • 48
    Clancy-Hepburn, K., Hickey, A. A., Nevill, G. (1974) Children's behavior responses to TV advertisements. J Nutr Edu. 6: 9396.
  • 49
    Perks, S. M., Roemmich, J. N., Sandow-Pajewski, M., et al (2000) Alterations in growth and body composition during puberty. IV. Energy intake estimated by the Youth-Adolescent Food-Frequency Questionnaire: validation by the doubly labeled water method. Am J Clin Nutr. 72: 14551460.
  • 50
    Krebs-Smith, S. M., Graubard, B. I., Kahle, L. L., Subar, A. F., Cleveland, L. E., Ballard-Barbash, R. (2000) Low energy reporters vs. others: a comparison of reported food intakes. Eur J Clin Nutr. 54: 281287.