|Home | About | Journals | Submit | Contact Us | Français|
The energy density (ED; kcal/g) of foods, when manipulated in the laboratory, affects short-term energy intake. The aim of this study was to examine if, when given a choice, dietary ED (foods only) and energy intake (expressed as a percentage of subjects’ estimated daily energy requirement) at a self-selected, single meal differs for teens born with a different familial predisposition to obesity and as a function of their sex. Subjects (13 males, 17 females) were 12 years of age and born at high-risk (HR; n = 15) or low-risk (LR; n = 15) for obesity based on maternal pre-pregnancy body mass index (BMI; kg/m2). The buffet meal, served for lunch and consumed ad libitum, consisted of a variety of foods and beverages with a range in ED. HR subjects consumed a more energy-dense meal (foods only) than LR subjects (1.84 vs. 1.42 kcal/g; P = 0.02) and males consumed a more energy-dense meal than females (1.83 vs. 1.43 kcal/g; P = 0.03). Total energy intake, when expressed as a percentage of subjects’ daily EER, did not differ between HR and LR subjects (42% vs. 33%; P = 0.16). Males, compared to females, consumed ~ 59% more energy from foods and beverages during the meal (46 vs. 29%; P = 0.008). During a single multi-item lunch meal, teens with a familial predisposition to obesity and males, independent of their obesity risk status, self-selected a more energy-dense meal. Familial risk for obesity, through either genetic or environmental pathways, may facilitate a more energy-dense diet.
The effects of energy density (ED; kcal/g) on energy intake have been systematically studied in the laboratory (Kral & Rolls, 2004; Rolls, 2000). The results of these studies have shown that as the ED of meals increased, so did energy intake both in children (J. O. Fisher, Liu, Birch, & Rolls, 2007; Leahy, Birch, & Rolls, 2008) and adults (Bell, Castellanos, Pelkman, Thorwart, & Rolls, 1998; Bell & Rolls, 2001; Kral, Roe, & Rolls, 2002; Rolls et al., 1999). Likewise, epidemiological studies (Poppitt & Prentice, 1996) and cross-sectional surveys (Ledikwe, Blanck, Kettel Khan et al., 2006; Ledikwe, Blanck, Khan et al., 2006; Mendoza, Drewnowski, Cheadle, & Christakis, 2006) have provided evidence that energy intake is affected by ED. A prolonged consumption of diets high in ED has been associated with increased weight status and weight gain in some, but not all studies (Drewnowski, Almiron-Roig, Marmonier, & Lluch, 2004; Kral et al., 2007; Stubbs, Johnstone, O’Reilly, Barton, & Reid, 1998). Behavioral interventions aimed at reducing dietary ED were beneficial for weight loss and weight management (Ello-Martin, Roe, Ledikwe, Beach, & Rolls, 2007; Rolls, Roe, Beach, & Kris-Etherton, 2005), thus supporting a causal link between ED and a positive energy balance.
Parental obesity, and mothers’ weight status in particular, have been shown to predict adult obesity among both obese and non-obese children (Whitaker, Wright, Pepe, Seidel, & Dietz, 1997). It is conceivable that mothers influence their children’s eating behaviors and weight development through genetic and environmental (e.g., modeling of eating behavior, feeding practices) pathways. To date little is known about the genetic and learned familial influences on the relationship between ED and energy intake (McCrory, Saltzman, Rolls, & Roberts, 2006). There is evidence in children for learned preferences for energy-dense over more energy-dilute foods (Birch, 1999). As with many behavioral traits, there likely exist individual differences among children in self-selected dietary ED. These differences may be due to inherited genetic differences, to differences in children’s early rearing environment, or other factors.
Moreover, there is evidence for genetic contributions to daily and meal-specific dietary ED (de Castro, 2006). From a sample of identical and fraternal adult twin pairs, estimated genetic effects explained between 37% and 50% of the variance in meal-specific and daily ED. Thus, it is conceivable that children may seek energy-dense foods by ways of a genetic predisposition to the taste or higher energy content of those foods.
On the other hand, home or other environmental factors such as the availability and accessibility of energy-dense foods have been shown to play an important role in the formation of children’s foods preferences and consumption patterns (Birch, 1999). For example, repeated exposure to foods was associated with an increased liking and consumption of these foods. Thus, parental feeding practices may enhance innate predispositions to the liking of more energy-dense foods in young children and may establish eating habits which favor an increased consumption of energy-dense foods. It is also possible that parents of obesity-prone children restrict children’s access to energy-dense foods in an effort to moderate their energy intake and weight gain. The use of overly restrictive feeding practices, however, is associated with greater eating in the absence of hunger (Birch, Fisher, & Davison, 2003; Francis & Birch, 2005) and increased child weight status (Faith, Scanlon, Birch, Francis, & Sherry, 2004).
The aim of this study was to assess whether familial predisposition to obesity and participants’ sex was associated with dietary ED and energy intake of 12-year-old adolescents during a self-selected single multi-item meal in the laboratory. We hypothesized that subjects born at high risk for obesity and males would self-select a more energy-dense meal and ingest more calories than would subjects born at low risk.
The 12 year-old subjects in this study were participants in an ongoing longitudinal investigation designed to examine the growth and development of children born with or without a familial predisposition to obesity. Subjects were enrolled at the age of 3 months and have undergone annual assessments. They were classified as being at high or low risk for obesity on the basis of their mothers’ pre-pregnancy body mass index (BMI; kg/m2). Parental obesity, and mothers’ weight status in particular, has been identified as a strong risk factor for obesity in the offspring (Whitaker et al., 1997). Comparing eating behaviors of children whose parents differ in obesity status is an accepted strategy for testing whether genetic vulnerability for obesity may express itself through eating traits (Faith, 2005). Mothers of the HR children in the initial cohort averaged a pre-pregnancy BMI of 31.2 kg/m2 and those of the LR children averaged 19.4 kg/m2. Details of parental characteristics and study design were reported previously (Stunkard, Berkowitz, Schoeller, Maislin, & Stallings, 2004; Stunkard, Berkowitz, Stallings, & Cater, 1999; Stunkard, Berkowitz, Stallings, & Schoeller, 1999).
All subjects were Caucasian and the assessment was scheduled ± 3 months of subjects’ 12th birthday. Subjects who participated in this study (N = 52) was a sub-sample from the larger cohort (N = 72) (Berkowitz, Stallings, Maislin, & Stunkard, 2005) in this longitudinal study and represented the youth who were available for their Year 12 assessment. Out of the 52 subjects who took part in the Year 12 assessment, 46 participated in the meal experiment. The remaining 6 subjects did not participate in the experimental meal procedure due to either dietary restrictions or because they were not feeling well that day.
In addition, after the first 16 subject visits there was a protocol change to better accommodate families’ schedules. As a consequence, a subset of subjects (n = 16) consumed the experimental meal as dinner; the remaining subjects (n = 30) consumed the meal as lunch. There were fundamental differences in the set-up of the experimental meal between these protocols. Most importantly, the timing of meals consumed prior to the buffet meal differed in the two protocols. In the initial protocol in which the experimental meal was served at dinner, participants fasted for 5 hours prior to the buffet meal, whereas participants in the revised protocol fasted for only 3 hours. Likely due to this difference in the timing of meals, there were differences in participants’ perceived hunger and prospective consumption before the buffet meal (see assessment procedure below) indicating that children who consumed the experimental meal at dinner were more hungry before the meal and could eat more during the meal compared to children who consumed the buffet meal for lunch. Due to these differences in the duration of fasting prior to the experimental meal, we chose to base the current analyses on the 30 subjects (13 males, 17 females) who consumed the experimental meal at lunch.
Written informed consent was obtained from the parents and assent was obtained from the youth. The study protocol was approved by the Institutional Review Boards of the University of Pennsylvania and The Children’s Hospital of Philadelphia (CHOP).
The buffet lunch meal was part of 1.5 day assessment period (including an overnight stay) at the CHOP Clinical and Translational Research Center. Subjects were admitted to the hospital on the evening before the test day. Once admitted, subjects were not allowed to leave the hospital until the completion of all assessments on the next day. On the day of their test meal, subjects completed a variety of anthropometric and body composition assessments.
On the morning of their test session, subjects were served a standard breakfast which they consumed ad libitum. Breakfast consisted of a choice of Cheerios (Fun Pack, 17 g), Frosted Flakes (Fun Pack, 34 g), Raisin Bran (Fun Pack, 43 g), plain bagel (~65 g), full-fat cream cheese (56 g), fat-free cream cheese (42 g), butter (10 g), jelly (28 g), banana (~120 g), hot water for tea (~220 g), sugar (10 g), artificial sweetener (4 g), orange juice (237 ml), 2% milk (473 ml), and cold water (296 ml). Breakfast for all subjects provided approximately 1,500 calories. Breakfast was completed by 9:00AM and subjects were instructed not to eat or drink anything (except water) until the test lunch meal 3 hours later.
The buffet meal consisted of a variety of foods (15 food items, 5 sauces and condiments) and beverages (7 beverage items) with a range in ED (Table 1). All food and beverage items were provided by the CHOP food service and from two national food chains including a family-style and a fast food restaurant. For all subjects, the foods and beverages were served in individual containers. Labels indicating the exact name of the foods were displayed with every food item. The buffet meal provided approximately 5200 calories. Subjects, who were alone in the room during lunch, could freely choose the types and amounts of foods and beverages they wanted to consume. Their lunch meal was video-taped. All foods and beverages were weighed (0.1 g) prior to serving, and reweighed after subjects finished eating, to determine the amount consumed. Nutritional information derived from manufacturer food labels and from the restaurants where the foods and beverages were purchased.
Immediately before and after each meal, subjects were asked to rate their perceived hunger, thirst, prospective consumption (how much food they thought they could eat), nausea, and fullness by using a 100-mm visual analog scale (VAS) with opposing anchors. Specifically, subjects were asked “How hungry are you right now?”, “How thirsty are you right now?”, “How much food do you think you could eat right now?”, “How nauseated do you feel right now?”, and “How full do you feel right now?”. The anchors for each question were labeled “not at all hungry/extremely hungry”, “not at all thirsty/extremely thirsty”, “nothing at all/a large amount”, “not at all nauseated/extremely nauseated”, and “not at all full/extremely full”, respectively. After the meal, subjects were also asked to rate how typical the amount that they ate during the meal was with anchors being marked “less than typical” and “more than typical.”
All data were analyzed using the SAS software (version 9.1, SAS Institute, Inc., Cary, NC). Independent-samples t-tests were used to compare subject characteristics (i.e., height, weight, BMI, BMI z-score, EER) by risk group. For all VAS ratings (i.e., hunger, thirst, prospective consumption, nausea, fullness, typical amount), nonparametric t-tests were used to compare means between risk groups. The computation of subjects’ estimated daily energy requirement was based on sex- and weight-specific equations (Medicine, 2005). The physical activity coefficient was set to be 1.00 (i.e., sedentary) for all subjects.
The main outcome variables were the ED and energy intake of the self-selected meal. The ED of the meal, which was based on participants’ intake, was computed by dividing total calories by total weight of food consumed at lunch. Computations of ED and resulting statistical analyses were carried out by excluding all beverages (Ledikwe et al., 2005). Energy intake was expressed as a percentage of subjects’ estimated daily energy requirement (EER) for energy consumed from foods, beverages, and foods and beverages combined. A secondary outcome variable was the amount (in gram) consumed from caloric and non-caloric beverages.
A 2 (sex) x 2 (risk group) general linear model multivariate ANOVA tested simultaneously the main effects of risk group and sex on the ED of the lunch meal, energy intake at lunch, and beverage intake. The interaction between risk group and sex was tested for significance in all models and removed if not significant. Analyses which tested the effects of risk group and sex on meal ED were conducted with and without adjusting for subjects’ BMI z-scores in the model.
Data are presented as model-based means ± SEM, the difference between the means, and the 95% confidence interval (CI) of the mean difference between the means. P-values less than 0.05 were considered statistically significant for all analyses.
As expected and previously reported (Berkowitz et al., 2005; Faith et al., 2006), anthropometric measures and daily EER differed between the HR and LR groups. HR subjects (6 males, 9 females), compared to LR subjects (7 males, 8 females), had a significantly higher BMI (23.2 ± 6.5 vs. 17.4 ± 2.1 kg/m2), BMI z-score (0.90 ± 1.17 vs. - 0.45 ± 0.88), and daily EER (1958 ± 290 vs. 1712 ± 135 kcal/day), respectively. Six HR subjects (40%) were obese (BMI-for-age ≥ 95th percentile), while all of the LR subjects were of normal weight (BMI-for-age 5 - 84th percentile).
Table 2 shows before- and after-meal ratings of perceived hunger, thirst, prospective consumption, nausea, and fullness as well as after-lunch rating of typicality of the amount of food consumed by risk group. None of the risk group differences were statistically significant (P > 0.06). There was considerable variability in subjects’ VAS ratings as indicated by measures of skewness and kurtosis.
When offered to self-select from a buffet meal, HR subjects, compared to LR subjects, ingested a meal which was ~ 30% more energy dense (Figure 1). This difference in meal ED (Diff: 0.43 kcal/g; 95% CI: 0.09, 0.77 kcal/g) remained borderline significant when controlling for participants’ BMI z-score (P = 0.045). Males, compared to females, ingested a meal which was ~ 27% more energy dense; a difference (Diff: 0.39 kcal/g; 95% CI: 0.05, 0.74 kcal/g) which was significant (P = 0.03). The difference in meal ED between males and females remained significant when controlling for participants’ BMI z-score in the model (P = 0.04).
Energy intake from foods, when expressed as a percentage of subjects’ daily EER, did not differ between HR and LR subjects (36 ± 4% vs. 29 ± 4%; Diff: 7; 95% CI: -4, 17%; P = 0.21). Males, compared to females, consumed ~ 44% more energy from food during the meal (39 ± 4% vs. 27 ± 3%; Diff: 12%; 95% CI: 1, 22%; P = 0.03).
Energy intake from beverages did not significantly differ between HR and LR subjects (6 ± 1% vs. 4 ± 1%; Diff: 2%; 95% CI: -1, 5%; P = 0.27). Males, compared to females, consumed ~ 133% more energy from beverages during the meal (7 ± 1% vs. 3 ± 1%; Diff: 4%; 95% CI: 1, 7%; P = 0.005).
There was a trend for HR subjects to consume more caloric beverages (322.0 ± 42.6g) than LR subjects (208.2 ± 42.6g; Diff: 113.8g; 95% CI: -11.9, 239.5g; P = 0.07). However, this trend was eliminated when adjusting for BMI z-score (P = 0.17). There was no significant difference between HR and LR subjects in the amount of non-caloric beverages they consumed (216 ± 35.1g vs. 267.0 ± 36.7g; Diff: 50.9g; 95% CI: -50.6, 152.5g; P = 0.28).
There also was no significant difference between males and females in the amount of caloric (293.9 ± 38.5g vs. 236.4 ± 47.1g; Diff: 57.5g; 95% CI: -70.8, 185.8g; P = 0.36) and non-caloric beverages (290.5 ± 51.4g vs. 192.5 ± 24.3g; Diff: 97.9g; 95%CI: -33.1, 229.0g; P = 0.12) they consumed.
Total energy intake did not differ between HR and LR subjects (42 ± 4% vs. 33 ± 4%; Diff: 8%; 95% CI: -3, 20%; P = 0.16). Males, compared to females, consumed ~ 59% more energy from foods and beverages during the meal (46 ± 4% vs. 29 ± 4%; Diff: 16%; 95% CI: 5, 28%; P = 0.008).
This study showed that, during a single multi-item meal, subjects who were born at HR for obesity self-selected a more energy-dense meal compared to subjects born at LR for obesity. This association remained significant when adjusting for current BMI z-score. Energy intake from food, beverages, or food and beverages combined, however, when expressed as a percentage of subjects’ daily energy requirement, did not differ between risk groups.
It is possible that familial influences, genetic or environmental in nature, may have affected HR subjects’ susceptibility to select foods of higher ED. For example, the home food environments in which HR and LR teens were raised may have been very different (Davison & Birch, 2001). In households with a family history of obesity teens may have had easier access to more energy-dense foods during their upbringing which in turn may have helped shape their preferences for those foods. Given that parents, and mothers in particular, serve as important role models for eating behavior and food choices (Brown & Ogden, 2004; Cutting, Fisher, Grimm-Thomas, & Birch, 1999; Faith, 2005; J. Fisher, Mitchell, Smiciklas-Wright, & Birch, 2001; Longbottom, Wrieden, & Pine, 2002; Vauthier, Lluch, Lecomte, Artur, & Herbeth, 1996), it furthermore is conceivable that mothers of youth born at HR for obesity may have passed on to their children some of their own food preferences. It is also possible that, during their upbringing, parents of HR youth may have restricted access to energy-dense foods more than parents of LR youth, which in turn may have enhanced HR teens’ preference for these types of foods.
Another interesting finding was that the association between meal ED and risk group was independent of subjects’ weight status. This suggests that teens’ food choices (with respect to ED) were different for HR subjects compared to LR subjects regardless of whether HR subjects were of normal-weight or overweight. If sustained, the susceptibility of HR subjects to select a more energy-dense diet may predispose these subjects to negative long-term health consequences, such as development of obesity and related chronic diseases. Cross-sectional data from adults have shown that diets that are higher in ED also tend to be of lower diet quality (Ledikwe, Blanck, Khan et al., 2006). Poor diet quality has been associated with adverse health consequences in U.S. adolescents such as the metabolic syndrome (Pan & Pratt, 2008).
Another interesting finding was that energy intake from foods, energy intake from beverages, and total energy intake, when expressed as a percentage of youths’ estimated daily energy requirement, did not significantly differ between HR and LR subjects. This finding, combined with the significant risk group effects for ED, suggests that HR and LR subjects may have consumed different amounts (by weight) of foods and beverages. Given that the weight of food per volume differed for the various food items (e.g., 1 cup of fruit salad and mixed salad weigh approximately 175g and 55g, respectively), it would have been necessary to develop a single, composite variable (e.g., z score) to standardize and hence compare the amount of food consumed (McConahy, Smiciklas-Wright, Birch, Mitchell, & Picciano, 2002). However, the small sample size precluded us from computing a composite variable and hence analyzing the weight of food that was consumed. Because the beverages that we served in this study had a more similar weight per volume, we were able to compare HR and LR subjects in the amounts of caloric and non-caloric beverages they consumed. The finding of a trend for HR subjects to consume more caloric beverages than LR subjects warrants additional research.
Our findings also point to important sex differences in intake during early adolescence. With respect for meal ED, there was a significant association between meal ED and sex such that males selected a more energy-dense meal than did females (model adjusted for risk group). Data from the Continuing Survey of Food Intakes by Individuals (CSFII, 1994 - 1996), which were based on two 24-h dietary recalls, showed significant sex differences in dietary ED among adults (> 19 years of age) (Ledikwe et al., 2005). In this analysis, women consumed a diet that was significantly lower in ED compared to that of men. On the other hand, when the CSFII (1994 - 1996, 1998) data were analyzed for children (mean age 9.3 years), no sex differences were found in dietary ED (Mendoza et al., 2006). It is possible that sex differences in dietary ED start to emerge during early puberty perhaps to support the higher energy needs for growth and development in males.
With respect to energy intake, males, independent of their risk status to obesity consumed an overall greater percentage of their daily energy requirement during lunch (from food, beverages, and food and beverages combined) than did females. It remains unclear what may have contributed to these differential intakes among males and females. It is possible that the laboratory conditions or other factors such as, for example, dieting behaviors or social desirability may have affected females’ and males’ eating behavior differently.
The strengths of this study include the precise measurement of teens’ intake under controlled laboratory conditions and use of an obesity risk design. Limitations are also associated with this experiment. The sample of teens was homogeneous with respect to race (i.e., all children were Caucasian) which precludes generalization of the findings to others. Second, this study cannot discern to what extent differences in the amounts of foods consumed may have contributed to differences in dietary ED between risk groups and males and females. Third, teens were videotaped during the lunch meal which could have affected both the amount of food consumed as well as their food choices. Fourth, the study used an estimated, rather than observed, measure of physical activity for the computation of participants’ EER. Finally, the study could not disentangle genetic from environmental influences on eating behavior phenotypes.
In summary, teens with a familial predisposition to obesity self-selected a more energy dense lunch meal than did teens without a familial predisposition. Future studies should determine the extent to which the consumption of higher energy-dense foods may foster normal growth patterns, particularly among males, or may instead represent one of the pathways involved in the development of obesity and related chronic diseases.
Source of support: NIH Grant DK068899 and the Clinical and Translational Research Center (Grant RR00240) of the Children’s Hospital of Philadelphia.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.