Caloric Compensation and Appetite Control in Children of Different Weight Status and Predisposition to Obesity

Abstract

To prevent childhood obesity it is critical to identify behavioral phenotypes for overeating, especially among children who are predisposed to obesity. We examined caloric compensation and appetite control in 212 normal-weight (NW) and obese (OB) children, ages 7 to 9, who were at high risk (HR) or low risk (LR) for obesity based on maternal obesity. In a within-subjects crossover design, children ate breakfast, lunch, dinner, and snacks in the laboratory once a week for two weeks. Children’s percentage compensation index (%COMPX) was computed at breakfast. Twenty-five minutes before breakfast, children received one of two compulsory preloads, which varied in energy density (ED) and caloric content [Low ED (LED): 1.00 kcal/g; 100 kcal; High ED (HED): 1.60 kcal/g; 160 kcal]. Children’s appetite was measured hourly using Visual Analog Scales, which were used to compute 3-hour post-prandial area under the curve (AUCs) after breakfast and the satiety quotient (SQ), which allows between-group comparisons of a fixed amount of a food potency to reduce appetite sensations per unit of intake. There were no significant differences in %COMPX, SQ, or AUC among LR-NW, HR-NW, and HR-OB children (P>0.10). SQs for Hunger and Prospective Consumption were higher and SQ for Fullness lower after consuming the LED compared to the HED preload (P<0.009). Further, the SQ and AUC for Desire to Eat and AUC for Prospective Consumption significantly predicted energy intake during the remainder of the day (P<0.03). In this study, HR-NW children did not differ from LR-NW or HR-OB children in their caloric compensation or appetite control. Foods with a high satiating effect may facilitate appetite control and help to moderate daily energy intake in all children, including at-risk children.

Clinical Trial Registration

This study was registered at clinicaltrials.gov as NCT02928874.

Introduction

One in 5 children between the ages of 6 and 11 years suffers from obesity (BMI-for-age ≥ 95th percentile) (Hales, Carroll, Fryar, & Ogden, 2017). Obesity during childhood and a family history of obesity are two of the strongest risk factors for adult obesity (Berkowitz, Stallings, Maislin, & Stunkard, 2005; Whitaker, Wright, Pepe, Seidel, & Dietz, 1997). Identifying modifiable behavioral phenotypes for overeating and excess weight gain is critical for the prevention of childhood obesity, especially in children at high risk (HR) for obesity based on a family history of obesity. It needs to be determined if obesogenic eating phenotypes are expressed in (still) normal-weight (NW), but at-risk children. In particular, the extent to which HR-NW children are phenotypically similar in their eating phenotypes to low-risk (LR)-NW children remains unknown. It is also possible that HR-NW children may have already adopted eating phenotypes that resemble more closely those of HR obese (OB; HR-OB) children. If HR-NW children exhibit obesogenic eating behaviors that may put them on a path to developing obesity, identifying these behaviors early will be important targets for intervention.

Experimental research has identified several possible behavioral phenotypes for childhood obesity (Fisher & Kral, 2008; French, Epstein, Jeffery, Blundell, & Wardle, 2012; Kral & Faith, 2009). One of these phenotypes is impaired caloric compensation, which is often measured by the percentage caloric compensation index (%COMPX). Caloric compensation, a measure of individual differences in satiety, refers to the adjustment in intake in response to changes in the energy density (ED; kcal/g) of a fixed amount of food (preload) (Birch & Fisher, 1997; Rolls, 2009). It is assessed by administering a preload and, after a predetermined time delay, measuring its effects on subsequent intake at a test meal (Rolls, 2009). While previous studies identified poor caloric compensation as a useful index for short-term (single meal) intake regulation (Birch & Deysher, 1985, 1986; Cecil et al., 2005; Faith et al., 2012; Johnson & Taylor-Holloway, 2006; Kral et al., 2012) and risk factor for elevated weight status (Brugailleres, Issanchou, Nicklaus, Chabanet, & Schwartz, 2019; Carnell, Benson, Gibson, Mais, & Warkentin, 2017; Kral et al., 2012), it has yet to be determined to what extent it may serve as a marker for longer-term (daily) energy intake. Findings from a study in a naturalistic setting by Ebbeling and colleagues (2004) showed that adolescents with overweight or obesity consumed significantly more calories during a fast food meal in a food court, both in absolute terms or relative to their estimated daily energy requirements, than normal-weight adolescents. Further, youth with overweight or obesity also showed significantly higher total daily energy intakes on days when they consumed fast food compared to days when they did not consume fast food which suggests that youth with a higher weight status, in particular, may not compensate for higher energy intakes from energy-dense meals at subsequent meals.

In addition, subjective measures of appetite (e.g., perceived hunger and fullness) are relevant to the regulation of short-term energy intake. The measurement of subjective motivation to eat yields greater insights into feeding behavior than assessing food intake alone (Stubbs et al., 2000). Individual differences in children’s subjective appetite control, however, are not fully elucidated and the extent to which they map on to objectively measures eating phenotypes is not well characterized. Appetite sensations such as perceived hunger, desire to eat, or fullness reflect both objective (unconditioned or physiological) and subjective (conditioned or learned) components of appetite control (Stubbs et al., 2000). Two measures of appetite control are the satiety quotient (SQ) and the post-meal area under the curve (AUC) of appetite sensations (Drapeau et al., 2005; Drapeau et al., 2007). The SQ assesses the individual-level satiating capacity of a preload and the extent to which it reduces subjective appetite sensations per unit of intake (e.g., kcal) (Green, Delargy, Joanes, & Blundell, 1997). In contrast to caloric compensation, which assesses actual adjustments (reduction or increase) in food intake following a preload, the SQ measures the extent to which a preload reduces subjective appetite sensations (McNeil et al., 2013) and therefore provides information about children’s perceived satiety. The post-meal AUC, on the other hand, is a summary indicator of perceived appetite sensations measured over time after a meal has ended. In a study with adults, both the SQ and the post-prandial AUC of fullness were inversely correlated with total energy intake and relative energy intake (i.e., total energy intake adjusted for each individual’s energy expenditure) over a 12-hour day (Drapeau et al., 2005), but comparable data in children are widely lacking. Data from two small samples of children, ages 7 to 12, indicated that a test meal’s macronutrient composition differentially affected children’s post-prandial AUC of hunger and fullness (Lomenick, Melguizo, Mitchell, Summar, & Anderson, 2009; Maffeis et al., 2010). For example, data by Maffeis and colleagues (2010) showed that appetite AUCs were significantly reduced in 10 prepubertal boys with obesity after a moderate-fat (27% energy from fat) than after a high-fat (52% energy from fat) meal. If identified as risk factors for increased energy intake, improving impaired caloric compensation and appetite control in at-risk children could serve as important intervention targets for the prevention of childhood obesity.

Children’s early home food environment has been shown to play a critical role in the development of eating behaviors (Anzman, Rollins, & Birch, 2010; Birch & Ventura, 2009). It is currently unknown if home food environments that provide easy access to highly palatable, energy-dense foods over time may impede children’s ability to regulate their energy intake and lead to an overconsumption of calories.

The primary aim of this laboratory study was to assess caloric compensation at breakfast, appetite control, and (daily) energy intake across preload conditions in LR-NW, HR-NW, and HR-OB children. We hypothesized that HR-NW and HR-OB children would show poorer caloric compensation and more impaired appetite control (i.e., subjective sensations of hunger, fullness, desire to eat, and prospective consumption) than LR-NW children. In an effort to examine the relationship between caloric compensation, appetite control (SQ, AUC), and daily energy intake, we further hypothesized that poor caloric compensation would be associated with higher daily energy intake and energy intake during the remainder of the day (sum of calories consumed at lunch, dinner, snacks, and at home) and impaired appetite control would be associated with higher breakfast and daily energy intake and higher energy intake during the remainder of the day. A secondary aim was to examine the association between children’s home food environment and their laboratory-assessed behavioral eating phenotypes. We hypothesized that children who live in home food environments that provide easy access to energy-dense foods will show impaired caloric compensation and poor appetite control.

SUBJECTS AND METHODS

Experimental Design

This study used a repeated-measures random order crossover design. Three distinct groups of children were enrolled in this study: 1) LR-NW: children of normal weight with a low risk for obesity; 2) HR-NW: children of normal weight with a high risk for obesity; and 3) HR-OB: children with overweight/obesity and a high risk for obesity. Children’s risk for obesity was based on their mothers’ current weight status. Children with mothers who had overweight or obesity were classified as at high risk of obesity; children with mothers who were normal-weight were classified as at low risk of obesity. LR-OB children were not enrolled because this study focuses on the prevention of childhood obesity by examining if HR-NW children’s eating phenotypes resemble more closely those of LR-NW or HR-OB children. Participants ate breakfast, lunch, dinner, and snacks in the laboratory one day a week for two weeks with a 7-day period between test visits. At each test visit, all children were served the same foods and beverages. Children consumed their meals and snacks in the same room, but room dividers prevented them from seeing each other to minimize the influence of social and visual cues on food intake. Participants’ percentage caloric compensation index (%COMPX) was assessed at breakfast using the preloading paradigm (Rolls & Hammer, 1995). Participants’ weight and risk status (LR-NW, HR-NW, HR-OB) were between-subjects factors and preload condition (LED, HED) was the within-subjects crossover factor. At 7 different time points throughout the morning of each test visit, participants’ perceived appetite sensations (i.e. hunger, fullness, prospective consumption) were assessed to calculate SQs and AUCs. Daily energy intake refers to the sum of calories consumed from all meals and snacks (including the preload) in the laboratory and at home (after the laboratory visit had ended).

Participants and Recruitment

Participants in this study were 212 racially/ethnically diverse boys and girls between 7 and 9 years of age and their mothers living in Philadelphia. The age range (7 to 9 years) was chosen to ensure that children had the cognitive ability to reliably complete the appetite ratings using Visual Analog Scales (VAS) (Beyer & Aradine, 1988; Erickson, 1990; Ginsburg & Opper, 1969) and to independently spend two full days in the laboratory. Furthermore, this age represents a critical period for obesity prevention as children have relatively greater autonomy to make food choices such that the capacity of their eating predisposition to shape obesity risk may be increasingly important. All mothers were children’s biological mother and primary caregiver (i.e., person responsible for daily care, grocery shopping, and child feeding). Families were recruited through newspaper and online advertisements, email advertisements sent to families in the healthcare network of the Children’s Hospital of Philadelphia (CHOP), poster displays at local public transportation stations, direct mass mailings, and flyers distributed in the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) offices and at local grocery stores.

A power analysis was conducted for the primary aim including %COMPX, daily energy intake, and the satiety quotient (SQ). Estimated group means and variances were derived from previous research conducted by Kral et al. (Kral et al., 2012; Kral, Whiteford, Heo, & Faith, 2011) and from the literature (Cecil et al., 2005; Faith et al., 2012; Fiorito, Marini, Francis, Smiciklas-Wright, & Birch, 2009; Johnson & Taylor-Holloway, 2006; Rodrigues, Leitao, & Lopes, 2013; Tybor, Lichtenstein, Dallal, Daniels, & Must, 2011). The common standard deviations used were: 106.9 for %COMPX, 860.4 kcal for daily energy intake, and 11.5 mm/kcal for the SQ. Based on these estimates a sample size of 225 children (75 children per group) would give 80% power to detect a mean between-group difference in %COMPX of 54.3% (effect size: 0.51), daily energy intake of 361.4 kcal (effect size: 0.42), and a SQ of 4.8 mm (effect size: 0.42).

Screening

Families who were interested in participating in the study underwent a telephone screening interview during which they provided information about their child’s age, sex, medical history, food preferences, height, and weight as well as their own height and weight. Families who pre-qualified from the telephone interviews were invited to come to the Center for a screening visit.

During the onsite screening visit, mothers and children were asked to provide voluntary consent (mothers) and assent (children) to participate in the study by signing the consent and assent forms. The study was approved by the Institutional Review Board (IRB) of the University of Pennsylvania. Children and mothers then had their height and weight measured and children’s liking and preference for all study foods, beverages, and snacks was assessed using a validated taste preference assessment (Birch & Sullivan, 1991). During the assessment, children rated their liking of each study food and beverage using a hedonic Likert-type scale with 5 cartoons ranging from a frowning to a smiling face (1=‘dislike very much’, 2=‘dislike a little’, 3=‘just OK’, 4=‘like a little, 5=‘like very much’). A picture book which depicted high-resolution color images of each food and beverage item accompanied the assessment. Children were asked to taste a small amount (<1 teaspoon) of the preload (oatmeal) and main entrées for each meal. When tasting the oatmeal, children were also asked to indicate whether or not they could consume the entire portion of oatmeal presented in the bowl, which represented the amount of oatmeal children were required to consume in full during test visits. Children were asked to rate the remaining study foods by viewing the image of the food in the picture book but without tasting the foods.

To be included, children had to: be regular breakfast eaters; be between 7 and 9 years of age; meet the weight / risk group criteria (LR-NW, HR-NW, HR-OB; see Assessment of Height and Weight); and like most of the foods that were served in the study in order to ensure that they would consume the study foods. When establishing the liking criteria, we applied more stringent liking criteria to foods that were compulsory (preload), part of the caloric compensation protocol (breakfast items), and foods that provided the majority of the calories at meals (main entrées). That is, only children who rated 1) the oatmeal as “Like a Little” or above, 2) the buffet breakfast as “Just OK” or above, 3) the main entrées at lunch and dinner as “Just OK” or above, 4) at least 1 of the 2 lunch and dinner side dishes as “Just OK” or above, and 5) at least 3 out of the 6 fruits and 3 out of the 6 sweet/salty snacks as “Just OK” or above were enrolled in the study.

Children were excluded from participation if they had: serious medical conditions known to affect food intake or body weight; any developmental, medical, or psychiatric conditions that might impact study compliance (e.g., attention deficit hyperactivity disorder, autism, emotional disorders); a learning disability or poor reading ability/comprehension; visual or auditory impairment; food allergies related to the foods served in this study or lactose intolerance; were taking medications that are known to affect appetite, food intake, or body weight; were underweight (BMI-for-age < 5^th percentile); did not regularly consume breakfast; did not like or could not eat the foods and beverages served in the study; were unable to consume the compulsory preload in full; or had mothers who were pregnant.

Out of a total of 1,708 calls received, 811 children pre-qualified from the phone screening and 568 children participated in the onsite screening visit. Out of 245 children who met inclusion criteria for the study, 212 children participated in the study.

Caloric Compensation

Participants’ caloric compensation was assessed during breakfast using the preloading paradigm (Rolls & Hammer, 1995). Specifically, 25 minutes before breakfast, participants were given one of two compulsory oatmeal preloads that varied in ED (1.00 or 1.60 kcal/g); see detailed description of experimental menus below. The rationale for choosing oatmeal as the breakfast preload was two-fold. One, in our previous research (Kral, Bannon, Chittams, & Moore, 2016) we used oatmeal at breakfast in a similar sample of children (8 to 10 years; 85% African American; 10% more than one race) and children showed a high preference for the oatmeal (98% indicated that they liked the oatmeal “very much” or “a little”). Second, we needed a preload which could be manipulated in energy density without substantially impacting the sensory properties of the two preload versions.

Assessing caloric compensation at the first meal of the day (breakfast) enabled us to control hunger and assess children’s subsequent intake over an entire day. The order of the preload conditions which subjects received on their first test visit was randomized across groups of children participating in the test visits together. The subsequent preload condition was counterbalanced on the initial assignment. Children who were rescheduled due to missing a visit received their preload conditions in the order that they were originally randomized to.

To maximize the likelihood of detecting caloric compensation (Roe, Thorwart, Pelkman, & Rolls, 1999), we formulated the preload size to correspond to approximately one third of children’s typical energy intake at breakfast (300 – 400 kcal), which we derived from currently ongoing and previous studies conducted by Kral (Kral et al., 2016; Kral et al., 2011) and from the literature (Morgan, Zabik, & Leveille, 1981; Nicklas, Bao, Webber, & Berenson, 1993). Twenty-five minutes after being given the preload, participants were served a multi-item breakfast (test meal; Table 1), which they consumed ad libitum. We modeled the timing of the preload after Rolls and colleagues (Rolls et al., 1991) who showed that a shorter interval between preload administration and the self-selection meal was optimal for caloric compensation.

Table 1:

Amounts of foods and beverages served during test visits

Foods and beverages served	Energy density (kcal/g)	Amount (g or pieces)	Energy (kcal)
Preload (compulsory)
Oatmeal1
Low energy density (LED)	1.00	100 gram	100
High energy density (HED)	1.60	100 gram	160
Breakfast (ad libitum)
Waffles2 (toasted)	2.71	4 pieces	379
Syrup3	2.58	100 gram	258
Cereal4 (ready-to-eat)	4.19	50 gram	210
Milk (2% fat)5	0.56	140 gram	78
Toaster pastry6 (strawberry)	3.85	1 piece	200
Peaches7 (drained)	0.83	150 gram	124
Orange juice8	0.46	342 gram	157
Total number of calories offered during meal:			1,406
Lunch
Pasta (cooked)9 w/tomato sauce10	1.30	800 gram	1040
Broccoli11 w/butter12 and salt13	0.43	150 gram	65
Milk (2% fat)5	0.56	342 gram	192
Total number of calories offered during meal:			1,297
Dinner
Chicken Nuggets14	2.82	15 pieces	719
Potato rounds (deep fried potatoes)15	1.86	30 pieces	533
Green beans16 w/butter12	0.32	150 gram	48
Ketchup17	1.18	100 gram	118
Milk (2% fat)5	0.56	342 gram	192
Total number of calories offered during meal:			1,610
Post-meal snacks
Fruit
Pineapple18 (drained)	0.57	120 gram	68
Peaches7 (drained)	0.83	120 gram	100
Pears19 (drained & sliced)	0.78	120 gram	94
Golden Delicious apple20 (sliced)	0.57	120 gram	68
Banana20 (sliced)	0.89	120 gram	107
Mandarin oranges21 (drained)	0.65	120 gram	78
Total number of calories offered during snack:			515
Sweet and savory snacks
Milk chocolate22	5.00	100 gram	500
Potato chips23	5.71	30 gram	171
Chocolate chip cookies24	4.85	10 pieces	534
Pretzels25	3.93	35 gram	138
Baked snack crackers26	4.67	70 gram	327
Sponge cake w/cream filling27 (1/2 inch pieces)	3.51	2 pieces	270
Total number of calories offered during snack:			1,940

¹Quaker Oats Quick 1-Minute, 100% Whole Grain, The Quaker Oats Company, Chicago, IL [Note: Prepared with milk, water, butter, and sugar or sugar substitutes]

²Eggo Waffles Homestyle, Kellogg Company, Battle Creek, MI [Note: Nutrition Facts Information changed 2/2017; new ED: 2.57 kcal/g; energy served: 360 kcal]

³Aunt Jemima Original Syrup, The Quaker Oats Company, Chicago, IL

⁴Cinnamon Toast Crunch Cereal, General Mills, Minneapolis, MN

⁵Reduced Fat Milk, 2%, Giant, Landover, MD

⁶Poptarts, Frosted Strawberry, Kellogg Company, Battle Creek, MI

⁷Yellow Cling Sliced Peaches in Heavy Syrup (undrained), Del Monte Foods, San Francisco, CA

⁸Orange Juice, Original, No Pulp, Tropicana Products, Inc., Bradentown, FL

⁹Rotelle Pasta, San Giorgio, New World Pasta Company, Harrisburg, PA

¹⁰Italian Sauce, Traditional, Prego, Campbell Soup Company, Camden, NJ

¹¹Broccoli, Petite Broccoli Florets, Hanover Foods Corporation, Hanover, PA

¹²Butter, Unsalted Sweet, Land O’Lakes, Inc., Arden Hills, MN

¹³Iodized Salt, Morton Salt, Inc., Chicago, IL

¹⁴Chicken Breast Nuggets, Whole Grain, Perdue, Salisbury, MD [Note: Nutrition Facts Information changed 8/2016; new ED: 2.53 kcal/g; energy served: 750 kcal; Nutrition Facts Information changed 5/2017: new ED: 2.50 kcal/g; energy served: 788 kcal]

¹⁵Tater Tots, Ore Ida, Heinz Company, Pittsburgh, PA [Note: Nutrition Facts Information changed 1/2018; new ED: 1.51 kcal/g; energy served: 433 kcal]

¹⁶Green Beans (undrained), Fresh Cut, Del Monte Foods, Walnut Creek, CA

¹⁷Tomato Ketchup, Heinz Company, Pittsburgh, PA

¹⁸Pineapple Chunks in 100% Pineapple Juice (undrained), Dole, Westlake Village, CA [Note: Nutrition Facts Information changed 6/2018; new ED: 0.49 kcal/g; energy served: 59 kcal]

¹⁹Bartlett Pear Halves in Heavy Syrup (undrained), Del Monte Foods, San Francisco, CA

²⁰National Nutrient Database for Standard Reference Legacy release, United States Department of Agriculture

²¹Mandarin Oranges in Light Syrup (undrained), Dole, Westlake Village, CA

²²M&M’s, Milk Chocolate, Mars, Inc., Hackettstown, NJ [Note: Nutrition Facts Information changed 8/2016; new ED: 4.76 kcal/g; energy served: 476 kcal]

²³Lay’s Classic Potato Chips, Frito-Lay, Inc., Plano, TX

²⁴Nabisco Chips Ahoy! Original Chocolate Chip Cookies, Mondelez Global, LLC, East Hanover, NJ

²⁵Rold Gold Tiny Twists Pretzels, Frito-Lay, Inc., Plano, TX

²⁶Goldfish Baked Snack Crackers, Cheddar, Pepperidge Farm, Inc., Norwalk, CT

²⁷Twinkies, Hostess Brands, LLC, Kansas City, MO [Note: Nutrition Facts Information changed 11/2016; new ED: 3.38 kcal/g; energy served: 260 kcal]

Participants’ caloric compensation ability was assessed by evaluating the percent caloric compensation index (%COMPX) (Cecil et al., 2005; Faith et al., 2012; Johnson & Birch, 1994; Johnson & Taylor-Holloway, 2006), which was computed as follows:

$$\text{\%COMPX}=\left[\left({\text{EI}}_{\text{LED}}–{\text{EI}}_{\text{HED}}\right)/\left({\text{Preload}}_{\text{HED}}–{\text{Preload}}_{\text{LED}}\right)\right]\text{x}100$$

where EI _{low ED} corresponds to calories consumed from the test meal after the LED preload, EI _{high ED} corresponds to calories consumed from the test meal after the HED preload, Preload _HED corresponds to calories consumed from the compulsory HED preload, and Preload _LED corresponds to calories consumed from the compulsory LED preload. A %COMPX of 100% equals accurate compensation; %COMPX < 100% indicates under-compensation (overeating); and %COMPX > 100% indicates over-compensation (undereating). The greater the deviation (above or below) from a COMPX of 100%, the more “impaired” an individual’s caloric compensation ability is thought to be. Consistent with previous research, in addition to calculating %COMPX, we also examined children’s mean energy intake at breakfast (test meal and preload) by preload condition across risk groups (Roe, Meengs, & Rolls, 2012; Williams, Roe, & Rolls, 2014).

Appetite Control

For the assessment of appetite control, children were asked to rate their perceived hunger (“How hungry do you feel right now?”), desire to eat (“How strong is your wish to eat right now?”), prospective consumption (“How much food do you think you could eat right now?”), and fullness (“How full do you feel right now?”) using 100-mm VAS with opposing anchors (e.g., “extremely hungry” or “not at all hungry”) and 8:35AM (before preload), 8:45AM (after preload), 9:00AM (before breakfast), 9:30AM (after breakfast), 10:30AM, 11:30AM, and 12:30PM (before lunch). VAS have been successfully used in children to assess perceptions of appetite (Anderson, Saravis, Schacher, Zlotkin, & Leiter, 1989; Bekem et al., 2005; Kral et al., 2011) and pain (B. J. Shields, Cohen, Harbeck-Weber, Powers, & Smith, 2003; B.J. Shields, Palermo, Powers, Grewe, & Smith, 2003). VAS ratings were used to calculate SQs and AUCs.

Satiety Quotient (SQ)

The satiating value of a fixed amount of food (preload) was evaluated with the SQ (Green et al., 1997). The satiety quotient allows between-group comparisons of a fixed amount of a food’s (preload) potency to reduce appetite sensations per unit of intake (kcal) at a specific timepoint after consumption. The SQ was calculated for each child, preload, VAS appetite rating (i.e. hunger, desire to eat, prospective consumption, fullness) as follows:

$$\text{SQ}\left(\text{mm/kcal}\right)=\left[\text{pre-preload}\text{fasting}\text{rating}–25\text{min}\text{post-preload}\text{rating}\right]/\text{energy}\text{content}\text{of}\text{preload}]\text{x}100$$

Thus, the SQ measures the degree to which individuals feel sated in response to a fixed amount of food (satiating efficiency). A higher SQ for hunger, desire to eat, and prospective consumption indicates a greater satiating value of the preload whereas a lower SQ would indicate a blunted satiating value of the preload. The reverse holds true for the SQ of fullness. The decision to use the 9:00am (or 25-min post-preload) appetite ratings to calculate SQs was based on Green and colleagues (Green et al., 1997) who established the SQ. We chose this time point because a) the measurement of the SQ requires abstinence of food or drink consumption during the period in which appetite ratings are recorded and 2) to allow time for physiological processing of nutrients from the preload.

Post-Prandial Area under the Curve (AUC) of Appetite Ratings

Children’s satiating value of a meal, consumed ad libitum, was evaluated by calculating a 3-hour AUC of each VAS appetite rating (i.e. hunger, desire to eat, prospective consumption, fullness) after breakfast. The post-meal AUC allows between-group comparisons of an ad libitum meal’s potency to reduce appetite sensations measured repeatedly over a discrete period of time after consumption. Consistent with previous research, the AUC for each time period (9:30AM – 10:30AM, 10:30AM – 11:30AM, 11:30AM – 12:30PM) was computed using the trapezoid method, which inscribes or circumscribes the number of trapezoids under the curve (Doucet, St-Pierre, Almeras, & Tremblay, 2003). The sum of the areas of the trapezoids yields the total AUC.

Experimental Menus

The ED (kcal/g), amounts (weight and volume), and energy contents (kcal) of foods and beverages served at breakfast, lunch, dinner, and snacks are shown in Table 1. The size of most servings fell between the 75^th and 90^th percentile of intake per eating occasion for children ages 6 to 11 years, based on data from the Continuing Survey of Food Intakes by Individuals (Smiciklas-Wright, Mitchell, Mickle, Goldman, & Cook, 2003). The total amount of energy provided during the test meals was more than children would likely consume, and portion sizes remained constant across experimental preload conditions.

Preload

Twenty-five minutes before breakfast, participants were asked to consume in full one of two oatmeal preloads that varied in ED (1.00 or 1.60 kcal/g). The HED preload provided 160kcal; the LED preload provided 100kcal. The amount of preload served in both conditions remained the same (100g). The oatmeal preloads were formulated using the Food Processor SQL software (Version 9.8; 2005; ESHA Research, Salem, OR). Variations in preload ED were achieved by modifying the types or amounts of sugar [regular sugar vs. no calorie sweetener (Splenda)], milk, water, and butter. The two preloads were similar in sensory properties. The percentage energy from protein, carbohydrates, and fat were 16%, 55%, and 29% for the LED preload and 13%, 57%, and 30% for the HED preload, respectively. Both preloads were served in an 8-oz Styrofoam bowl.

Breakfast

The breakfast included waffles with syrup, ready-to-eat cereal with 2% milk, peaches, toaster pastry, and orange juice. Portion sizes remained constant across experimental preload conditions. The waffles were served on a 10.25-inch diameter plate, the syrup was served in a 3.25-oz transparent soufflé cup, the cereal and peaches were served in 12-oz bowls, the 2% milk was served in a 5-oz pitcher, the toaster pastry was served on a 7-inch diameter plate, and the orange juice was served in a 14-oz translucent cup with a lid and straw. This breakfast served as the test meal on which the computation of %COMPX was based.

Lunch

Lunch consisted of pasta with tomato sauce, broccoli, and 2% milk. During the meal preparation, we added a small amount of unsalted butter (8g or 1/2 tablespoon) and salt (0.5g or 0.09 teaspoons) per 400g broccoli to enhance the broccoli’s palatability. The pasta with tomato sauce was served on a 10.25-inch diameter plate, the broccoli was served in a 12-oz bowl, and the milk was served in a 14-oz translucent cup with a lid and straw.

Dinner

Dinner consisted of chicken nuggets, potato rounds, ketchup, green beans, and 2% milk. During the meal preparation, we added a small amount of unsalted butter (8g or 1/2 tablespoon) per 400g green beans to enhance the beans’ palatability. The green beans were canned and already contained sea salt. The chicken nuggets and potato rounds were served on a 10.25-inch diameter plate, the ketchup was served in a 3.25-oz transparent soufflé cup, the green beans were served in a 12-oz bowl, and the milk was served in a 14-oz translucent cup with a lid and straw.

Snacks

Children were served a variety of fresh fruit (i.e., pineapple, peaches, pears, apple, banana, mandarin oranges) or sweet and savory foods (i.e., chocolate, potato chips, chocolate chip cookies, pretzels, crackers, cake) as snacks offered after lunch and dinner. All snacks were served to participants in individual 12-oz bowls. The after-meal snacks were offered to assess eating in the absence of hunger (EAH), a second behavioral eating phenotype which is described in Kral et al. (under review).

All experimental meals, preloads, and snacks were prepared in the research kitchen of the Center by trained staff according to a standardized protocol. All foods and beverages were weighed before being served to the participants and reweighed after they finished eating to determine the amount consumed by each child to the nearest 0.1g.

Food records

After each test visit, children’s intake during the remainder of the day (from the time children left the laboratory after dinner until they went to bed) was determined using telephone-administered dietary recalls. At the end of each test visit, mothers were provided with a food record to help them keep track of the types and amounts of foods and beverages their children consumed during the remainder of the day. They also received detailed instructions on how to accurately report their children’s intake and were provided with visual aids for portion size estimates. The visual aids provided examples of food serving sizes and corresponding everyday objects as visuals to assist with portion size estimation (e.g., one medium fruit was depicted as about the size of a tennis ball). Trained research staff called mothers on the morning following each test visit to obtain information about the types and amounts of foods their child ate during the prior evening. Data from the dietary recalls were entered and analyzed using the Food Processor SQL software (ESHA Research, Salem, OR).

Daily energy intake

For the purpose of this analysis, daily energy intake refers to the sum of calories consumed from all meals and snacks (including the preload) in the laboratory and at home (after the laboratory visit had ended); energy intake during the remainder of the day refers to the sum of calories consumed at lunch, dinner, snacks, and at home. The estimated energy requirement (EER) for sedentary boys and girls between 7–9 years ranges between 1400 – 1600 kcal/day and 1200 – 1400 kcal/day, respectively (United States Department of Agriculture, 2015).

Assessment of Child and Maternal Height and Weight

At the screening visit, children’s and mothers’ height and weight were measured by trained research staff. All measures were taken with families’ wearing light clothing and having their shoes removed. Weight was measured on a digital scale (Tanita BWB-800, Arlington Heights, IL; accurate to 0.1 kg) and standing height was measured on a wall-mounted stadiometer (Veder-Root, Elizabethtown, NC; accurate to 0.1cm). Measurements were recorded in duplicate; the intra-person mean was used for statistical analyses. Child age- and sex-specific BMI percentiles and z-scores were calculated using the Center for Disease Control and Prevention Growth Charts 2000⁶². Children were classified as normal-weight (BMI-for-age 5 – 84^th percentile) or having overweight/obesity (OB; BMI-for-age ≥ 85^th percentile) (Ogden et al., 2002). Maternal BMI was computed as weight (kg) divided by height (m) squared. Mothers were classified as normal-weight (BMI 18.5 – 24.9 kg/m²), having overweight (BMI 25.0 – 29.9 kg/m²), or having obesity (BMI ≥ 30 kg/m²) (Expert Panel on the Identification, Evaluation, and Treatment of Overweight in Adults, 1998).

Home Food Environment

Families’ home food environment was assessed by the Home Food Inventory (Fulkerson et al., 2008). This inventory consists of 190 foods that represent 13 major food categories. Mothers were asked to complete the inventory while at home, by inspecting all areas of the home where food is stored, including the refrigerator, freezer, pantry, cupboard, and other areas, such as the basement). Food items are listed in a checklist type format with yes/no (1/0) response options with higher scores indicating greater availability. The inventory differentiates between regular-fat and reduced-fat and regular sugar and low sugar items and categorizes foods into healthful and less healthful groupings based on the food’s fat and sugar content. The inventory produces an obesogenic household food availability score (ranging from 0–71), which represents the summative score for regular-fat versions of cheese, milk, yogurt, other dairy, frozen desserts, prepared desserts, savory snacks, added fats; regular-sugar beverages; processed meat; high-fat quick, microwavable foods; candy; access to unhealthy foods in refrigerator and kitchen. This score was used in the statistical analysis. The inventory has been validated (Fulkerson et al., 2008) and has shown acceptable criterion and construct validity for all major food category scores and the obesogenic home availability score.

Procedures

Children’s eating behaviors were assessed in the laboratory during two test visits during which children were required to stay at the Center from 8:00 AM until 5:30PM. Before the start of the actual test visits, participants attended an orientation session during which they met the other children in their group and were trained on how to complete VAS ratings and ate a dinner meal to get acclimated to the laboratory environment. Children came to the laboratory in groups of 1–4. On the day of each test visit, mothers were instructed to have their children refrain from eating and drinking (except water) from the time they got up in the morning until the start of the visit at 8:00 AM. Upon arriving for the test visit, mothers were asked to complete a brief report to ensure that they complied with these instructions.

At each visit, participants were served the same meals and snacks. Twenty-five minutes before breakfast they were asked to consume one of two compulsory oatmeal preloads which varied in ED (1.00 or 1.60 kcal/g) over a 10-minute period. After a 15-minute delay, participants were served a multi-item breakfast at 9:00AM, which they consumed ad libitum over a period of 25 minutes. Lunch and dinner were served at 12:30PM and 4:30PM, respectively, and were also consumed ad libitum over a period of 25 minutes. At each visit, children were served either fruit or sweet and savory snacks after lunch (1:15PM) and after dinner (5:15PM). Children were instructed to eat as much or as little as they liked during the meals and snacks. Children were able to drink water up to one hour before lunch and dinner if desired. Research staff remained in the laboratory during the entire visit to ensure that the children adhered to study instructions. At 14 time points throughout the test visit, children completed VAS to assess their appetite. During times when no formal assessments were taking place, research staff engaged children in games, watch movies, or complete homework assignments. Mothers waited in a room nearby or in close proximity to the Center for the duration of the test visits. Figure 1 outlines the timeline for the assessments during a full day visit.

Figure 1:

Timeline of assessments during full-day test visit.

Statistical Analysis

Data were analyzed using the SAS System for Windows (Version 9.4; SAS Institute, Cary, NC). We used the Shapiro-Wilk test in conjunction with distribution plots and summary statistics to confirm normality of distribution of continuous variables. Mixed linear model (PROC MIXED) analysis with repeated measures was used to compare participants groups (LR-NW, HR-NW, and HR-OB) in %COMPX, appetite control (i.e., SQs and AUCs for hunger, prospective consumption, desire to eat, fullness), daily energy intake (sum of calories consumed at all meals and snacks (including the preload) and at home), and energy intake during the remainder of the day (sum of calories consumed at lunch, dinner, snacks, and at home) across preload conditions. Fixed factor effects in all mixed linear models were preload condition (LED, HED), participant group, and test visit (i.e., test visit 1 or 2). An unstructured covariance matrix was used to adjust for lack of independence among the repeated measures within participants. The interaction between preload condition and risk group was tested for significance and removed if not significant. We further performed non-parametric 1) Wilcoxon Signed Rank tests to determine whether participants’ %COMPX differed from 0 or 100 and 2) Spearman rank correlation analysis to assess the relationship between %COMPX and participants’ BMI z-scores. Non-parametric tests were used for the full sample because the data contained an outlying %COMPX value. We also tested for sex differences in %COMPX using the non-parametric Kruskal-Wallis test. A mixed model analysis, which included preload condition, risk group, test visit, and either %COMPX, SQs, or AUCs for hunger, prospective consumption, desire to eat, or fullness in the model, was used to assess whether %COMPX or appetite control significantly predicted daily energy intake or energy intake during the remainder of the day. For the analysis of the secondary aim, a mixed model analysis, which included preload condition, risk group, test visit, and the obesogenic home availability score in the model, was used to assess whether participants’ home food environment significantly predicted %COMPX, SQ, and post-prandial AUCs.

For each outcome we also conducted a sensitivity analysis which excluded 1) 18 participants who did not consume at least 90% of the preload (12 participants did not meet the intake requirement during one test visit; and 6 participants did not meet the intake requirements during both test visits) and 2) 42 participants who had unusual VAS ratings (n = 42). The determination of “unusual” VAS ratings were based on Hunger and Fullness ratings. Hunger scores of 0 to 10 (low hunger) and Fullness scores of 90 to 100 (extremely full) immediately before a meal were considered to be unusual. Hunger scores of 90 to 100 (extremely hungry) and Fullness scores of 0 to 10 (low fullness) immediately after a meal were also considered to be unusual. Participants with less than 80% of “usual” Hunger and Fullness VAS were excluded in the sensitivity analysis. In the sensitivity analysis, we tested for possible effects of child race and household income by including these variables as covariates in analyses related to the main aims. We also compared the variance heterogeneity in participants’ %COMPX among risk groups using Levene’s test for homogeneity of variance. Lastly, we performed t-tests to determine whether participants’ %COMPX differed from 0 or 100 and Pearson correlation analysis to assess the relationship between %COMPX and participants’ BMI z-scores. These tests were used in the sensitivity analysis because data were normally distributed and contained no outlying %COMPX values. We also re-tested for sex differences in %COMPX using Analysis of Variance (ANOVA). Results from the sensitivity analyses can be found in the Supplementary Materials.

Descriptive statistics for demographic and anthropometric characteristics and the obesogenic household food availability score are reported as means (± SDs) for continuous variables or as percentages for categorical variables. Results from the mixed linear model analysis are presented as model-based means (± SEM). Reported P values are 2-sided and P < 0.05 was considered significant for all tests.

Results

Demographic and Anthropometric Characteristics and the Obesogenic Home Availability Score

Table 2 depicts the demographic and anthropometric characteristics of children and their mothers as well as the obesogenic home availability score. There were no significant differences between groups in age, sex, and ethnicity (P < 0.132). Groups did significantly differ in race with a higher proportion of HR-NW and HR-OB children being Black or African American (∼81%) and a higher proportion of LR-NW children being White (27%) and more than one race (29%). HR-OB children, when compared to HR-NW and LR-NW children, had a significantly higher weight and height (P < 0.0001). All groups significantly differed in BMI-for-age percentiles and BMI z-scores (P <.0001). The mean BMI z-score among LR-NW children was significantly lower than 0 (Mean = - 0.24, t-test P = 0.008); the mean BMI z-score among HR-NW children was significantly greater than 0 (Mean = 0.16, t-test P = 0.028); and the mean BMI z-score among HR-OB children was significantly greater than 0 (Mean = 1.75, t-test P < 0.0001).

Table 2:

Subject demographic and anthropometric characteristics and the home food environment by group

Characteristic	LR-NW (N = 60)	HR-NW (N = 77)	HR-OB (N = 75)	P-value
Children
Age (years), mean ± SD	8.3 ± 0.66	8.3 ± 0.84	8.5 ± 0.84	0.349
Sex, n (%)
Male	28 (46.7%)	29 (37.7%)	29 (38.7%)
Female	32 (53.3%)	48 (62.3%)	46 (61.3%)	0.520
Race, n (%)
American Indian	0 (0%)	1 (1.3%)	0 (0%)
Asian	3 (5.1%)	1 (1.3%)	0 (0%)
Native Hawaiian/Pacific Islander	0 (0%)	0 (0%)	0 (0%)
Black or African American	23 (39.0%)	61 (79.2%)	62 (82.7%)
White	16 (27.1%)	2 (2.6%)	3 (4.0%)
More than one race	17 (28.8%)	11 (14.3%)	10 (13.3%)
Unknown	0 (0%)	1 (1.3%)	0 (0%)	<0.001a
Ethnicity, n (%)
Hispanic	8 (17.0%)	5 (7.6%)	3 (4.9%)
Not Hispanic	35 (74.5%)	59 (89.4%)	55 (90.2%)
Unknown	4 (8.5%)	2 (3.0%)	3 (4.9%)	0.132
Height (cm), mean ± SD	131.2 ± 6.61	131.0 ± 7.11	137.5 ± 7.64	<0.001
Weight (kg), mean ± SD	27.2 ± 4.01	28.4 ± 4.57	42.6 ± 10.29	<0.001
BMI z-score, mean ± SD	−0.2 ± 0.69	0.2 ± 0.63	1.7 ± 0.46	<0.001
BMI-for-age percentile, mean ± SD	42.5 ± 22.57	56.3 ± 22.12	94.5 ± 4.14	<0.001
Weight group, n (%)
Normal-weight	60 (100%)b	77 (100%)	0 (0%)
Overweight/obese	0 (0%)	0 (0%)	75 (100%)	<0.001
Mothers
Age (years), mean ± SD	36.9 ± 6.78	33.1 ± 6.32	36.2 ± 6.56	0.001
Academic degree, n (%)
High School	24 (42.9%)	50 (69.4%)	41 (63.1%)
College	14 (25.0%)	18 (25.0%)	14 (21.5%)
Master’s	12 (21.4%)	4 (5.6%)	10 (15.4%)
Doctorate	6 (10.7%)	0 (0%)	0 (0%)	<0.001
Household income, n (%)
Less than $25,000	23 (39.0%)	38 (50.7%)	34 (45.9%)
$25,000 – $50,000	10 (16.9%)	25 (33.3%)	28 (37.8%)
Greater than $50,000	26 (44.1%)	12 (16.0%)	12 (16.2%)	0.001
Marital status, n (%)
Single	25 (42.4%)	51 (68.0%)	53 (70.7%)
Married	26 (44.1%)	15 (20.0%)	17 (22.7%)
Divorced, separated, widowed	8 (13.6%)	9 (12.0%)	5 (6.7%)	0.005
Obesogenic household food availability score	28.0 ± 12.2	30.8 ± 12.3	26.9 ± 10.8	0.120

^aCategories used for covariate analysis: White/Other, Black/African American, and more than one race

^bOne child was underweight at the time of study enrollment

Mothers significantly differed in age, race/ethnicity, academic degree, household income, and marital status across groups (P < 0.001). HR-NW mothers were significantly younger than HR-OB and LR-NW mothers. A higher proportion of HR-NW and HR-OB mothers were Black or African American (∼83%) and a higher proportion of LR-NW mothers were White (39%) and Hispanic (18%). A higher proportion of LR-NW mothers had advanced degrees (college or above; 57%) when compared to HR-NW (31%) and HR-OB (37%) mothers. A higher proportion of LR-NW mothers (44%) had household incomes of greater than $50,000 compared to HR-NW and HR-OB mothers (16%). A higher proportion of LR-NW mothers were married (44%) when compared to HR-NW and HR-OB mothers (∼21%). Lastly, there was no significant between-group difference in the obesogenic household food availability score by risk group (P = 0.120).

Percentage Caloric Compensation Index (%COMPX)

There was no statistically significant between-group difference in %COMPX (P = 0.86). The mean %COMPX for LR-NW, HR-NW, and HR-OB children was 54.6 ± 48.9, 39.9 ± 43.7, and 73.9 ± 43.4, respectively. There was no significant correlation between %COMPX and participants’ BMI z-score (Spearman rank correlation coefficient: r = −0.06; P = 0.39). Further, %COMPX was not significantly associated with participants’ sex. The %COMPX also did not significantly predict daily energy intake (P = 0.34) or energy consumed during the remainder of the day (P = 0.61). The mean %COMPX was 56.3, which was significantly greater than 0 (Wilkoxon P = 0.03) and significantly lower than 100 (Wilcoxon P = 0.003).

Energy Intake during Preload – Test Meal Paradigm

Figure 2 depicts energy intake at breakfast (test meal and preload) by preload condition across risk groups. There was a significant main effect of risk group (P = 0.0021) indicating that total energy intake at breakfast was significantly higher in HR-OB children (713.6 ± 26.2 kcal) than in LR-NW (586.9 ± 29.3 kcal) and HR-NW (610.6 ± 25.9 kcal) children. Furthermore, there was a significant main effect of preload condition (P = 0.0009) indicating that children consumed, on average, 37 more calories at breakfast in the HED (656 ± 17 kcal) than in the LED (618 ± 17 kcal) preload condition. The main effect for visit was also significant (P = 0.02) indicating that children consumed, on average, 39 more calories during the first visit (657 ± 18 kcal) when compared to the second visit (617 ± 16 kcal), respectively.

Figure 2:

Model-based means (± SEM) of energy intake during Preload – Test Meal Paradigm for LR-NW (n = 60), HR-NW (n = 77), and HR-OB (n = 75) children. LED, low energy density; HED, high energy density. The models for these means were adjusted for time and preload condition. Main effects were significant for risk group (P = 0.0021), preload condition (P = 0.0009), and test visit (P = 0.02). The risk group-by-preload condition interaction was not significant (P = 0.95).

Daily Energy Intake

With respect to daily energy intake, there was a significant main effect of risk group (P < 0.0001) showing that daily energy intake significantly differed between all groups (LR-NW: 2238 ± 72 kcal; HR-NW: 2510 ± 63 kcal; HR-OB: 2808 ± 64 kcal; Figure 3). The main effect for test visit was also significant (P = 0.02) indicating that children consumed, on average, 89 more calories during test visit 1 (2563 ± 44 kcal) when compared to test visit 2 (2496 ± 42 kcal), respectively. There was no significant risk group-by-preload condition interaction (P = 0.10) or main effect of preload condition (P = 0.21).

Figure 3:

Model-based means (± SEM) of daily energy intake for LR-NW (n = 60), HR-NW (n = 77), and HR-OB (n = 75) children.

Appetite Control

The model-based means (± SEM) for VAS, SQs, and AUCs for Hunger, Desire to Eat, Prospective Consumption, and Fullness can be found in Figure 4.

Figure 4:

Model-based means (± SEM) for the visual analog scale (VAS) appetite ratings, post-preload satiety quotient (SQ), and post-breakfast 3-hour area under the curve (AUC) for LR-NW, HR-NW, and HR-OB children. *Pairwise comparisons for significant risk group-by-time interaction (P < 0.04); ^a HR-NW > HR-OB; ^b HR-NW > LR-NW; ^c HR-OB > LR-NW

Satiety Quotient (SQ)

There were no statistically significant differences in children’s SQs for Hunger, Desire to Eat, Prospective Consumption, and Fullness across groups (P > 0. 10). The results did, however, show that children’s SQs for Hunger, Prospective Consumption, and Fullness, but not Desire to Eat, significantly differed by preload condition. Specifically, children showed a significantly higher SQ for Hunger (17.7 ± 2.2 vs. 9.1 ± 1.4; P = 0.0003) and Prospective Consumption (12.2 ± 2.1 vs. 6.5 ± 1.3; P = 0.008) and a lower SQ for Fullness (−15.6 ± 2.5 vs. −8.2 ± 1.4; P = 0.009) after consuming the LED when compared to the HED preload. This suggests that the LED preload immediately following consumption reduced hunger and prospective consumption and increased fullness more per unit of energy than the HED preload.

The SQ for Hunger (P = 0.005), Desire (P = 0.01), and Prospective Consumption (P = 0.02), but not Fullness (P = 0.63), significantly predicted energy intake at breakfast. That is, for each unit (mm/kcal) increase in SQ for Hunger, Desire to Eat, and Prospective Consumption, energy intake at breakfast decreased by 0.85, 0.81, and 0.77 calories, respectively. The interactions between the SQ dimensions and preload condition were not significant for energy intake at breakfast (P > 0.31).

Only the SQ for Desire to Eat significantly predicted daily energy intake (P = 0.001) and Remainder (P = 0.03) in all children.

Area Under the Curve (AUC)

There were no statistically significant differences in children’s AUCs for Hunger, Desire to Eat, Prospective Consumption, and Fullness across groups (P > 0.42). There also was no significant main effect of preload condition (P > 0.35) or risk group-by-preload condition interaction (P > 0.10) on appetite AUCs.

The breakfast AUCs for Desire to Eat, but not Hunger (P = 0.24), Fullness (P = 0.32), or Prospective Consumption (P = 0.05), significantly predicted daily energy intake (P = 0.03) in all children. The AUCs for Desire to Eat and Prospective Consumption, but not Hunger (P = 0.08) or Fullness (P = 0.18), significantly predicted energy intake during the remainder of the day (P < 0.02) in all children.

Relationship between Obesogenic Household Food Availability Score and Outcome Variables

The obesogenic household food availability score did not significantly predict %COMPX (P = 0.19), appetite SQs (all Ps > 0.18), or appetite AUCs (all Ps > 0.16).

Discussion

In this study, we found that obesogenic eating phenotypes related to caloric compensation and appetite control did not manifest themselves in children at healthy weight but high risk for obesity. Measures of appetite control (SQ, AUC) for select appetite sensations, but not %COMPX, significantly predicted energy intake at breakfast and energy intake over the course of the day. Specifically, children’s SQ for Hunger, Desire to Eat, and Prospective Consumption significantly predicted energy intake at breakfast, but only children’s SQ and AUC for Desire to Eat significantly predicted daily energy intake. Results further showed that the LED preload reduced hunger and prospective consumption and increased fullness more per unit of energy than the HED preload. This suggests that ingestion of a LED preload before a meal can help improve subjective appetite control and lead to a concomitant reduction in energy intake at the subsequent meal in children irrespective of their weight status and predisposition to obesity. Lastly, in this study the obesogenic household food availability score, a measure of the availability of high-fat, sugar and/or processed foods in the home, was not associated with children’s %COMPX or appetite control.

This study aimed to compare LR-NW, HR-NW, and HR-OB children in their caloric compensation ability. We had hypothesized that children at high risk for obesity (i.e., HR-NW, HR-OB) would show a poorer ability to compensate for the calories from the preload when compared to children at low risk for obesity (LR-NW). Contrary to what we had hypothesized, groups did not significantly differ in %COMPX including the variability of %COMPX. In both the full analysis and the sensitivity analysis, participants in all three groups undercompensated (COMPX < 100%). Some previous studies (Carnell et al., 2017; Faith et al., 2012; Kral et al., 2012), but not all (Cecil et al., 2005; Remy, Issanchou, Chabanet, Boggio, & Nicklaus, 2015), showed that children with overweight or obesity had an impaired ability to compensate for calories from a preload. There are several factors that may account for the non-significant between-group difference in %COMPX in our study. One, while we aimed for the preloads to provide approximately one third of children’s typical energy intake at breakfast (Roe et al., 1999), the actual percentage of calories provided by the preload was smaller (LED: ∼15%; HED: ∼24% for groups combined). Participants in this study showed larger energy intakes at the test meal (breakfast) than anticipated which may in part have been driven by the multi-item breakfast. Greater food variety, and with that greater exposure to sensory stimuli, has been shown to increase food and energy intake at meals in children and adults (Norton, Anderson, & Hetherington, 2006; Raynor & Vadiveloo, 2018; Roe, Meengs, Birch, & Rolls, 2013). Other factors that may be underlying the variability in findings when compared to other preload studies include, but are not limited to, the timing of the interval between preload and test meal, the form (semi-liquid) of the preload, and the size and composition of the test meal (Blundell et al., 2010; Rolls et al., 1991; Williams et al., 2014). The high variability in prior caloric compensation studies in children (Birch & Deysher, 1985, 1986; Carnell et al., 2017; Cecil et al., 2005; Faith et al., 2004; Faith et al., 2012; Kral et al., 2012) with respect to the type (solid, liquid, semi-solid), size, and ED/caloric difference of the preloads used complicates the comparison of study findings. Similar to ours, prior research (Carnell et al., 2017; Faith et al., 2004; Johnson & Birch, 1994) found substantial variability in %COMPX between individuals, but the underlying mechanisms by which predict more accurate compensation is achieved remain to be determined.

Another aim of this study was to assess appetite control in LR-NW, HR-NW, and HR-OB children. Our findings showed no significant differences between groups in SQ, AUC, and, when adjusted for multiple comparisons, VAS across four dimensions of appetite. This again suggests that children at normal weight and children with overweight or obesity who have a family history of obesity do not differ from normal-weight children without a family history of obesity in their appetite control as reflected in the satiating capacity of the preload and postprandial appetite sensations measured over time. We did, however, find a significant main effect of preload condition on SQ indicating that children showed a significantly higher SQ for Hunger and Prospective Consumption and a lower SQ for Fullness shortly after consuming the LED when compared to the HED preload. This indicates that, immediately following consumption, the LED preload reduced hunger and prospective consumption and increased fullness more per unit of energy than the HED preload. The HED preload contained 25% more oats and 22% less liquid (milk/water). Thus, it is likely that the higher water content, thinner texture, and modestly higher volume and protein content (16% of energy vs. 13% of energy) of the LED preload may have resulted in a higher satiating capacity (SQ) than the HED preload, which is consistent with other studies (Bertenshaw, Lluch, & Yeomans, 2013; Rolls, Bell, & Thorwart, 1999; Rolls et al., 1998). We further found that each unit increase in SQ for these appetite dimensions was accompanied by a concomitant decrease energy consumed at the subsequent breakfast. Together, these findings suggest that the LED preload, despite being lower in calories, exerted a greater satiating capacity than the HED preload and reduced caloric intake at the subsequent meal. While some (Rolls et al., 1999; Rolls et al., 1998), but not all (Birch & Deysher, 1985; Birch, McPhee, Steinberg, & Sullivan, 1990; Birch, McPhee, & Sullivan, 1989), prior preload studies in children and adults also documented a more pronounced reduction in energy intake from the test meal after consuming a LED preload, our findings are novel in that they suggest, for the first time, that LED preloads may exert their intake-suppressing effects through changes in children’s subjective appetite sensations.

Lastly, this study also identified SQs and AUCs for select appetite sensations (i.e., Desire to Eat and Prospective Consumption), but not %COMPX, as significant predictors of remaining and daily energy intake. This suggests that subjective measures of appetite control have implications for daily energy intake and foods with a high satiating efficiency between meals may help moderate daily energy intake via appetite control among children.

With respect to the study’s secondary aim, children’s home food environment, which was assessed by the obesogenic household food availability score, did not significantly predict caloric compensation or appetite SQs and AUCs. It is important to note that the Home Food Inventory only assesses availability of less healthy foods in the home, but it does not assess children’s actual consumption of these foods. It will therefore be important for future research to conduct a refined assessment of children’s food choices and intake outside of the laboratory and to relate them to behavioral phenotypes. It is also possible that eating phenotypes such as caloric compensation and appetite control are food- and situation-specific and may vary depending on the types of foods and beverages that are being consumed.

The strengths of this study include its large sample size and the highly controlled laboratory environment in which eating phenotypes were assessed. To our knowledge, this also is the first study that assessed caloric compensation ability in the context of overall appetite control in a unique cohort of children with different weight status and obesity risk. The study also had limitations. One, while the sample size of this study was large, we fell short of meeting our recruitment goal (93%). It is therefore possible that our study was not adequately powered to detect between-group differences in behavioral outcomes. Second, while VAS have been validated for use in children ages 7 and older and participants in our study received extensive training on how to complete VAS, some children may have had difficulty comprehending the scales or may have experienced fatigue with completing VAS at 14 different timepoints. Third, it is possible that the relatively small difference in ED and overall caloric content of the preloads may have accounted for not finding significant between-group differences in %COMPX. Also, while we attempted to match the two preload versions in palatability, it is possible that small differences in the taste and texture of the preloads may have led to differential appetizing effects and differentially impacted subsequent intake at the buffet breakfast. Fourth, children’s risk for obesity was based on maternal BMI and did not take into account fathers’ weight status. Given that both mothers’ and fathers’ weights contribute to the intergenerational transmission of BMI (Naess, Holmen, Langaas, Bjorngaard, & Kvaloy, 2016), the complete risk for obesity of the children who participated in this study is unknown. Lastly, the narrowly defined study sample of predominantly minority children may preclude a generalization of study findings to a more heterogenous group of children.

In summary, children who are normal weight but at risk of developing obesity based on a family history of obesity show a similar ability to compensate for calories at meals and perceived appetite control when compared children with a different weight status or family obesity risk. Therefore, being able to identify at-risk children based on these two eating phenotypes alone will be challenging. Perhaps a more comprehensive evaluation is needed which goes beyond the study of self-regulatory systems, such as satiety and appetite control, and includes appetitive traits reflecting food approach and motivation, such as sensitivity to reward or eating in the absence of hunger, to identify at-risk children early.

Supplementary Material

2

Supplemental Figure 1: Boxplot analysis of percentage compensation index (%COMPX) across risk groups (LR-NW: n = 49; HR-NW: n = 72; HR-OB: n = 69) from the sensitivity analysis. Open circles on the plot represent values that are more than 1.5 times the interquartile range (IQR) above the third quartile or below the first quartile.

Abbreviations

BMI

body mass index

%COMPX

percentage compensation index

VAS

visual analog scale

AUC

area under the curve

SQ

satiety quotient

LED

low energy density

HED

high energy density

NW

normal-weight

OB

obese

HR

high-risk

LR

low-risk