点击显示 收起
1 From the Bute Medical School, University of St Andrews, St Andrews, United Kingdom (JEC); the Biomedical Research Centre (CNAP) and the Centre for Public Health Nutrition Research (WW and IM), Ninewells Hospital and Medical School, University of Dundee, Dundee, United Kingdom; the MRC Human Nutrition Unit, Cambridge, United Kingdom (CB-S); the Chelsea School, University of Brighton, Brighton, United Kingdom (PW); and the School of Psychology, University of Liverpool, Liverpool, United Kingdom (DJW and MMH)
2 Supported by the UK Biotechnology and Biological Sciences Research Council (D13460). 3 Address reprint requests to MM Hetherington, Department of Psychology, Glasgow Caledonian University, Cowcaddens Road, Glasgow G4 0BA, United Kingdom. E-mail: marion.hetherington{at}gcal.ac.uk.
ABSTRACT
Background: Young children accurately compensate for energy-dense preloads consumed before test meals. The accuracy of compensation seems to deteriorate as a function of age.
Objective: The hypothesis that accurate energy compensation varies by age, body mass index, and individual characteristics of children and their mothers was tested.
Design: Energy intake (EI) from a test meal was measured in 74 children aged 6–9 y 90 min after the ingestion of no-energy (NE), low-energy (LE), or high-energy (HE) preload snacks. The NE preload consisted of 250 mL water, the LE preload consisted of a 56-g muffin + a 250-mL orange drink (783 kJ), and the HE preload consisted of a 56-g muffin + a 250-mL orange drink (1628 kJ).
Results: A significant dose-related reduction in EI was found after the preloads; younger children adjusted more effectively than did older children, although total EI (including preload energy) indicated that the adjustment was not accurate. The compensation index (COMPX) differed by preload and age group; COMPX scores were higher between the NE and LE preloads (younger children: 44.4 ± 9.3%; older children: 57.0 ± 11.6%) than between the NE and HE preloads (39.6 ± 4.9%; 31.3 ± 6.2%) and the LE and HE preloads (35.2 ± 7.8%; 7.4 ± 9.8%). This finding indicates a more consistent response across preloads and a greater sensitivity to energy load by younger than by older children. High interindividual variation and low intraindividual variation in COMPX was found. The tendency to over- or undereat in response to the preloads (deviation from perfect) correlated directly and positively with maternal concerns about child overweight, not with actual BMI.
Conclusions: The children adjusted their EIs in response to different preloads, and the younger children did so more effectively than did the older children. Poor short-term energy compensation may constitute a behavioral marker for positive energy balance.
Key Words: Children • eating behavior • energy compensation • food intake • child overweight
INTRODUCTION
Childhood obesity has doubled over the past 20 y (1, 2), and, although there is a general risk of excess weight gain from an increasingly obesigenic environment (3), it is critical to identify behavioral attributes that contribute to the risk of obesity early in development. Given the association between childhood and adult obesity (4–6), early intervention to prevent excessive weight gain (7) is warranted. Positive energy balance, achieved by consuming more energy than is expended, is key to the development of childhood obesity (8). Understanding controls of energy intake in children within and beyond single meals may provide markers for overconsumption in children.
Preschool children appear to be sensitive to the energy density of snacks given before a meal. In several studies, young children were shown to compensate accurately for the energy content of a snack given before lunch (9, 10). A compensation index (COMPX) was calculated for each child by dividing the difference in energy intake after 2 preloads by the difference in the energy content of the preloads, transformed to a percentage (11). A COMPX score of 100% reflects perfect compensation. Young children are generally good at compensating for high-energy snacks and can achieve COMPX scores between 50% (12) and 80% (10). In contrast with children aged <5 y, older children appear to compensate less effectively for high-energy snacks and have COMPX scores of 20% (10), which are more similar to adult scores (13). Indeed, some studies have shown no difference in compensation ability between 4- and 6-y-old children, adults, and older adults (14). Poorer self-regulation is associated both with increased adiposity of the child (11, 15) and with more controlling feeding styles of parents as determined by the Child Feeding Questionnaire (11). Clearly, many biological and behavioral factors influence short-term compensation, and these are worthy of further exploration.
Large interindividual variation in COMPX scores have been observed across different studies (14) and, within families, siblings fail to show similar compensation abilities (16). This suggests that some children are more sensitive to the energy content of preloads than are others. However, it is not clear whether there are intraindividual variations in self-regulation according to the type of preload used because many studies compare intakes on just 2 occasions to generate a single COMPX score.
The present study was designed to examine the response to preloads at different energy intakes (no energy, low energy, and high energy) and to investigate how the adjustment for preloads occurs relative to a control (water). It was hypothesized that younger children would have higher COMPX scores than would older children and that the ability to compensate would be stable over different preloads. Intraindividual variations in self-regulation were examined to identify children who are consistently good or poor regulators and to characterize features of the child or their mothers that might be associated with these differences.
SUBJECTS AND METHODS
Participants
Children aged 6–9 y from primary schools in and around the Tayside region of northeast Scotland, initially recruited as part of a larger study on the maintenance of energy balance, were invited to participate in this experiment. Children were allocated to groups by age; children aged 6–7 y and 8 mo constituted the "younger" group, and those aged 8–9 y and 8 mo constituted the "older" group. Thirty-seven boys (n = 19 aged 6–7 y and n = 18 aged 8–9 y) and 37 girls (n = 26 aged 6–7 y and n = 11 aged 8–9 y) participated in the study. Ethical approval for the overall study was granted by the Tayside Committee on Medical Research Ethics, the Fife Local Research Committee, and the Education Departments within each school authority. In addition, head teachers from each specific school approved the study before the parents were approached. The parents then provided written informed consent for their children’s participation.
Procedures
Children who normally consumed breakfast were invited to participate, and their parents were telephoned the evening before the test session to remind them to provide their children with their habitual breakfast the next morning. At school, children were asked what they had for breakfast that morning, and this was noted informally to check compliance. These measures were designed to increase compliance with the instruction to maintain habitual intake. Children were then asked to consume either a no-energy (NE1), low-energy (LE), or high-energy (HE) preload on 3 occasions midmorning, followed by a test-meal lunch 90 min later. This time profile replicated the children’s usual snack time at school followed by the test meal at the typical lunchtime. The order of preload administration was partially randomized; the water control preload was always given as the first condition. This served 2 purposes: to act as a control and to familiarize the children with the procedure and research assistants. All tests were separated by 1 wk and were conducted on a day when there was no physical education. As a check on order effects, a fourth water preload (NE2) was administered to 50% of the children. Data analyses were conducted for 3 preloads, because the test-meal intake during the second control preload did not differ from that during the first control.
The heights and body weights of the children were measured at school on the morning of the first test session. The heights and body weights of the mothers were measured at home or in the laboratory. Standing height without shoes was measured to the nearest 0.1 cm with a stadiometer (SECA, Bolton, United Kingdom). Body weight was measured to the nearest 0.1 kg with a mechanical floor scale (SECA) while the subjects were wearing light clothing. Body mass index (BMI) for each child and parent was calculated as weight (kg)/height2 (m).
Parents also completed a number of questionnaires, including the Child Eating Behavior Questionnaire (CEBQ; 17) and the Child Feeding Questionnaire (CFQ; 18). The 35-item CEBQ measures aspects of children’s eating styles according to 8 factors that measure the responsiveness to and enjoyment of food, satiety responsiveness, slowness in eating, fussiness, desire to drink, emotional overeating, and undereating. The internal consistency for these 8 factors is high and ranges from 0.74 for satiety responsiveness to 0.91 for enjoyment of food (17). Similarly, the 31-item CFQ measures parental attitudes toward, and strategies in, child feeding through 7 factors: perceived responsibility for child feeding, perceived parent weight, perceived child weight, concern about child weight, pressure to eat, restriction, and monitoring. Cronbach’s , which is used to assess internal consistency, is also high for these factors and ranges from 0.70 for pressure to eat to 0.92 for monitoring. More detailed psychometric properties of these tools were characterized and published elsewhere (17, 18). Items from the CFQ and CEBQ were correlated with COMPX to examine potential associations between these factors and short-term energy compensation.
Preloads and test meals
The nutrient contents of the 3 preloads are shown in Table 1. The 3 preloads were developed to differ in energy density. The NE control (0 kJ) consisted of 250 mL water, and the LE and HE preloads consisted of an orange drink and a muffin manipulated to differ only in total energy content. The LE preload (187 kcal, or 782.78 kJ) consisted of 250 mL orange drink (200 mL water + 50 mL low-energy orange diluting drink) and 56 g low-energy-dense muffins, and the HE preload (389 kcal, or 1628.35 kJ) consisted of 250 mL orange drink (200 mL water + 50 mL regular orange diluting drink) with the addition of 15 g maltodextrin (Maxijul; SHS International Ltd, Liverpool, United Kingdom) and 56 g regular-energy-dense muffins.
View this table:
TABLE 1. Energy, protein, fat, and carbohydrate contents of the preload foods
The NE control (water) was used to provide a baseline intake for comparison. Test meal intake was expected to be greatest after ingestion of the NE control, given that it had no energy density and weighed less than the other preloads. Energy density and the weight and volume of a food load are important factors in determining postingestive appetite and food intake (19, 20). These preloads were chosen to be familiar to the children and to incorporate a relatively large difference in energy content (202 kcal, or 846 kJ) with minimal differences in sensory properties. Systematic assessments were made to evaluate the pleasantness and likability of the orange drinks before the study began to ensure acceptance by the children. Each child was required to ingest 100% of the preload on each occasion.
The test meal was a self-selected lunch consisting of a variety of cold food items (Table 2) that were served to each child on individual trays. The lunch provided 7.95 MJ (1900 kcal) energy, which was much more energy than the children normally consumed at lunch, and was devised to offer sufficient quantity and variety to maximize choice. The children were allowed to request additional servings. A maximum of 30 min was allowed for the children to eat their chosen food items, and they were notified of the amount of time remaining 10 and 5 min before the 30-min period was reached. The average group size on each occasion was 4, and the children sat together during the lunch as they normally would at school. Research assistants were present to provide assistance and were instructed to lead any conversation away from the topic of food and eating.
View this table:
TABLE 2. Energy, protein, fat, and carbohydrate contents of foods and drinks available to children in the test meal
Assessment of intake
Energy intake at the test meal was assessed by weighing the food items before and after lunch and then using manufacturer’s information to calculate the total amount of energy consumed. The weights and energy contents of individual food items and drinks were calculated to assess food choice. The precision of energy compensation was assessed by using the COMPX, which was calculated as the difference in energy intake from the test-meal lunch on any 2 occasions divided by the difference in the energy content of those preloads. This value was converted to a percentage [(change in energy intake at the test meal/change between preload energy content) x 100] (11) A score of 100% indicates precise (calorie for calorie) compensation. Values less that 100% reflect undercompensation, values >100% reflect overcompensation. Total energy intake was calculated by adding energy from the preload to the energy intake at lunch.
Statistics
Intake data were analyzed by using repeated-measures analysis of variance, with age group and sex as between-groups factors. When significant main effects were obtained, post hoc tests with a Bonferroni correction factor for multiple comparisons were applied to determine the source of significant effects. Pearson’s correlation coefficient (r) was used to correlate between BMI, energy intake, and COMPX values. Spearman’s correlation coefficients (rs) were applied to correlations between intake and questionnaire scores. Deviation scores were calculated to assess how far children deviated from perfect compensation giving an absolute value. This provided a measure of the extent to which children over- or underconsumed at lunch after the preloads. Thus, the greater the deviation from perfect compensation, the higher the absolute (deviation) value. These deviation scores were correlated with child and maternal characteristics, including BMI and items from the CFQ and CEBQ. Cronbach’s was calculated for subscales on the CFQ and CEBQ to compare against published norms. Data were analyzed by using SPSS for WINDOWS (version 11.5; SPSS Inc, Chicago, IL), and the results are expressed as means ± SEMs unless otherwise stated. Statistical significance was set at P < 0.05.
RESULTS
A summary of the ages, body weights, heights, and BMIs of the children and their mothers and the number of overweight and obese children is provided in Table 3. Data were collected from 37 girls and 37 boys, of whom 15% were overweight and 8% were obese on the basis of age- and sex-appropriate international standards (21).
View this table:
TABLE 3. Age, height, weight, and BMI of the study cohort, by age group
Energy intake
Overall, the children adjusted their intakes at lunch in response to the energy content of the preload (P < 0.0001; Figure 1). As expected, the older children generally consumed more than did the younger children (2976 ± 119 kJ compared with 2534 ± 96 kJ; P < 0.001). The children’s energy intakes decreased linearly as the preload energy content increased (within-subjects contrast for linearity: P < 0.001). Post hoc tests showed that energy intake was significantly different between conditions. Thus, on average, the children ate significantly less after the HE preload than after the LE preload and less after both of these preloads than after the NE (water) preload (P < 0.001 in all cases; Table 4). No main effects of sex on energy intake were found when adjusted for body weight, so all analyses considered girls and boys together.
FIGURE 1.. Mean (±SEM) total energy intake (energy intake at lunch + energy intake from preload) after the no-energy (NE), low-energy (LE), and high-energy (HE) preloads. ANOVA for main effect of preload on energy intake at lunch: P < 0.0001; ANOVA for main effect of preload on total energy intake: P < 0.0001.
View this table:
TABLE 4. Energy intake at lunch, energy intake adjusted for body weight, and compensation index1
Total energy intake (including energy from the preloads) varied according to condition in a linear way (ANOVA: P < 0.0001), which indicated that the adjustment in food intake at lunch after the different preloads failed to accommodate precisely the energy content of the preloads (Figure 1). Thus, the total energy intake (including preload energy) was higher after the HE preload than after the LE and NE preloads.
Given that differences in energy intake emerged by age group, and that this difference was related to body size, energy intake was transformed into energy consumed as a function of body weight (kJ/kg). An interesting interaction emerged between preload and age group (ANOVA: P < 0.02), which suggested that the slope of adjustment between the NE and HE preloads was steeper for the younger than for the older children (Table 4). Thus, after energy intake as a function of body weight was accounted for, the younger children more effectively adjusted their energy intakes after the preloads than did the older children.
To test whether the changes in test-meal intake were specific to the preload energy and not due to increasing familiarity with the procedure or other temporal anomaly, we subjected 42 individuals to a fourth test, repeating the water-alone preload. Importantly, no changes in energy intake were observed in these individuals. Intake did not differ significantly between the 2 sessions (3038 ± 132 and 3054 ± 124 kJ), and energy intake in the 2 sessions was highly correlated (r(41) = 0.77, P = 0.001). Thus, the children ate the same amounts at lunch when given the water preload, and this provided an appropriate baseline against which to judge compensation after preloads that differed in energy content.
Compensation ability
The COMPX scores, which reflected the children’s ability to compensate accurately for the energy content of the preloads, confirmed the observation that, although adjustment occurred, few children managed perfect compensation (COMPX = 100%; n = 3). On average, with age groups combined, the best level of compensation (51%) was observed between the NE and LE preloads, and this compared with 35% and 21% for the NE/HE and LE/HE conditions, respectively. Compensation differed according to the type of preload eaten (ANOVA: P = 0.001). Post hoc tests showed that the ability to compensate accurately was higher between the LE and HE preloads and the control (water) than between the 2 energy-dense preloads (Table 4). Furthermore, a significant linear decline in COMPX score was found, which indicated that compensation accuracy followed this order: NE/LE > NE/HE > LE/HE (Table 4).
The interaction between preload and age group was significant (ANOVA: P = 0.04). It appeared that the younger children (40 ± 5%) tended to have higher overall COMPX scores than did the older children (32 ± 6%), but the difference was significant only between the LE and HE preloads (P < 0.05; Table 4). Within-groups analyses indicated that the COMPX score differed significantly by preload only for the older children; thus, COMPX was not significantly different between preloads in the younger children and varied significantly in the older children, ie, younger children were consistent in their response across preloads.
To assess the extent to which there were intraindividual differences in the ability to compensate between preloads, COMPX scores were correlated across preloads independently by age group. Overall, intraindividual variation was low, because correlations between COMPX scores by age group showed significant associations in the younger children [NE/LE:NE/HE (r(44) = 0.54, P < 0.001) and NE/HE:LE/HE (r(44) = 0.49, P < 0.001)] and older children [NE/LE:NE/HE (r(28) = 0.72, P < 0.001) and NE/HE:LE/HE (r(28) = 0.57, P < 0.001)]. COMPX scores were also highly correlated for the sample of 42 children who repeated the control condition. Thus, for COMPX NE/LE [NE1 = 53 ± 10%; NE2 = 55 ± 10%; r(41) = 0.38, P < 0.01] and COMPX NE/HE [NE1 = 36 ± 6%; NE2 = 37 ± 4%; r(41) = 0.44, P < 0.01] correlations were positive and significant. Clearly, COMPX was higher after the LE preload than after the HE preload compared with both control (NE) conditions (ANOVA: P < 0.05).
High positive correlations between COMPX scores after the NE preload suggests that compensation for an energy-dense preload occurs regardless of whether the preload is high or low in energy. However, the low COMPX score indicates that, generally, the adjustment was not accurate.
Food choice
To test whether intake of the preloads had a specific or general effect on food choice at lunch, macronutrient intakes and choice of individual food items were analyzed after each of the 3 preloads. Intakes in weight (g) of protein, fat, and carbohydrate decreased significantly as preload energy intakes increased. Post hoc comparisons confirmed that intake decreased in a linear manner with preload content. Children consumed the same percentage of energy as protein across preloads (13 ± 0.5%), but consumed significantly less energy as fat at lunch after the HE preload (37 ± 1%) than after the NE preload (40 ± 1%; P < 0.01) and a greater percentage of energy as carbohydrate after the HE preload (47 ± 1.0%) than after the NE preload (45 ± 1%; P = 0.053). Thus, the children’s energy intakes tended to be lower after the HE preload than after the NE preload, achieved by switching from sources of fat to sources of carbohydrate.
An evaluation of intakes of individual food items indicated that children tended to eat less of both types of cheese, bread, crackers, and raisins but ate the same amount of other foods in all preloads (ham, potato chips, chocolate, grapes, and orange juice). This pattern of intake might reflect preferences for these latter foods, which are not displaced by consuming the preloads.
Correlations
To identify characteristics of either the children or their mothers, which might predict the ability to compensate, a deviation score was calculated to assess the extent to which children deviated from 100% (perfect) compensation. This was then correlated with a range of variables, including the children’s BMI, the mother’s BMI, and scores on the CFQ and the CEBQ.
The BMI of the children was significantly correlated with maternal BMI (r(73) = 0.34, P < 0.01) and with total energy intake after all preloads (P < 0.01 in each case). The deviation score for NE/LE (
DISCUSSION
Overall, children in the present study adjusted their food intakes at lunch in relation to the energy content of the preloads administered. Thus, after a midmorning snack consisting of 783 or 1628 kJ, the children’s intake at lunch decreased 13% and 18%, respectively, compared with intake after water. Younger children adjusted their intakes in direct response to the energy content of the preload. However, an age difference in intakes indicated that older children generally consumed more than did the younger children, which would be expected. Thus, after body weight was accounted for, it was shown that the younger children adjusted their intakes directly in line with the energy content of the preload; they consumed 12% less after the LE preload and 22% less after the HE preload than did the older children, who reduced their intakes by 13% after the LE preload and by 15% after the HE preload. Clearly, younger children were more able to discriminate between the LE and HE preloads, which suggests that they were more sensitive to the energy content of the preloads than were the older children. Therefore, the younger children did not simply eat less after a snack, as expected, but were able to modify their intakes according to the energy content of the snack.
Despite these adjustments in intake at lunch after the midmorning snacks, intakes fell short of accurate compensation. Therefore, on days when the HE preload was consumed, total energy intake was significantly greater (by 34%). Thus, the consumption of a midmorning snack at school promoted energy intake and may have contributed to a positive energy balance.
The average compensation index of 30% indicated that children tended to adjust their intakes but not sufficiently to produce accurate compensation. This finding is in contrast with that of a recent study, which reported average compensation indexes of 103% between low-energy (12.55 kJ) and high-energy (627.6 kJ) preload drinks (14). Nevertheless, in both studies, average compensation masked better accuracy in some of the children than in the others. Indeed, Faith et al (16) showed large individual variation and no familial aggregation in COMPX scores between siblings. In the present study, the younger children tended to compensate for the differences in energy intakes better than did the older children. This finding supports previous research on short-term energy compensation from our laboratory (10).
It was noted that, despite high levels of interindividual variation, the ability to compensate within individuals correlated across preloads. This finding suggests that compensation accuracy is trait-related. This was supported when the control condition was repeated, because compensation scores were strongly correlated. Thus, the individual ability to compensate holds for different preloads. The poorest COMPX occurred between the LE and HE preloads compared with after either of these preloads and water. This would be expected because the task of differentiating between a water and an energy-dense preload is easier than discriminating between 2 preloads that share sensory and other properties. Clearly, children responded to the preloads by reducing their food intake at lunch, but they found it difficult to account for the differences in energy contents of the LE and HE preloads. However, there were large intraindividual variations; some children adjusted their intakes well across all preloads.
When the deviation from perfect compensation was correlated with individual and maternal characteristics, concern about the child’s weight status correlated with deviation scores. Indeed, greater concern about overweight was related to many characteristics of eating style associated with overconsumption, ie, faster eating, emotional overeating, food responsiveness, and lack of satiety. Therefore, although there was no direct correspondence between BMI and COMPX in children, concerns about overweight correlated both with deviation from perfect compensation and with markers of overeating. Taken together, these features could form part of a behavioral phenotype that carries an increased risk of overeating and obesity later in life. It is not clear whether a short-term measure of energy compensation relates to future obesity development, but it is possible that repeated exposure to energy-dense snacks, which are not compensated for, could lead to positive energy balance over time, and may provide a marker of overeating.
Previous research has shown that a child’s body weight status is related to self-regulation, with higher levels of adiposity associated with poorer compensation (11). The present study showed that maternal concern for a child’s weight is associated with self-regulation rather than with current BMI. It is difficult to ascertain whether concern for child weight could be a response to poor regulation in the child or whether concern influences poor regulation. However, it is interesting to note that concern for child weight has been shown to be associated with variance in adiposity and explains 15% of the variance in fat mass in children (22). Again, the direction of influence was unclear; nevertheless, concern for child weight is a modifiable behavior that could be tackled in childhood obesity-prevention programs.
Research with young children (ages 3–4 y) has shown that, with some training, the compensation index can be increased from 23% to 65% (23). This study used strategies to increase children’s self-regulation and to recognize internal cues of hunger and satiety. These findings could be extrapolated to interventions with older children, who seem less able to compensate accurately in the short-term for energy-dense snacks.
Food intake was adjusted at lunch after the energy-dense preloads, and this was achieved by reducing the percentage of energy as fat after the high-energy preload compared with the low-energy preload. In particular, children reduced their intakes of cheese, crackers, bread, and raisins, but their intakes of potato chips, chocolate, grapes, and orange juice remained the same after both energy-dense preloads. This pattern of choice is not easily understood in relation to sensory-specific satiety (24), which predicts that fewer foods that share sensory properties with the preload will be eaten, ie, orange juice and other sweet food items. It seems that foods that are highly favored by children are eaten in similar amounts, despite the intake of preloads moderately high in fat and carbohydrate. Thus, sweet snacks consumed midmorning did not displace sweet foods at lunch and were added to daily energy intake with the potential for contributing to overconsumption, if not compensated for later in the day.
Our previous research reported that preschool children adjusted their intakes at lunch after a sucrose preload by consuming fewer carbohydrates (10). The present investigation failed to support this finding and suggests that this precise regulation deteriorates in older children. Nevertheless, a recent study showed that children aged 9–12 y adjusted their food intakes at lunch according to the glycemic index of breakfast (25). Thus, children ate less at lunch after eating a breakfast with a low glycemic index than after a breakfasts with a higher glycemic index. Higher palatability and lower satiety experienced by children after the consumption of foods with a high glycemic index failed to reduce intakes at lunch. These observations may further our understanding of energy compensation in the short-term by highlighting the role of glycemic index, pleasantness, and satiety value on subsequent intake.
Future studies would benefit from ascertaining trait and context-related factors that predict compensation ability. It is clear, that there is large individual variation in compensation ability, and some children demonstrate precise compensation regardless of age. Poorer compensation has been associated with increased child adiposity (11) and greater parental control of child feeding (11), but these factors explain only part of the variance. It is possible that genetic variation may also account for some of the unexplained individual variation in energy compensation and in eating behavior in general, according to research conducted in animals and humans (26, 27). Further studies could facilitate the identification of children most at risk of positive energy balance, through both the transmission of risk from parents and from poor short-term energy regulation.
ACKNOWLEDGMENTS
We thank the parents and children who participated in this study.
CB-S, MMH, CNAP, WW, and PW wrote the proposal submitted to the Biotechnology and Biological Sciences Research Council. JEC, IM, and DJW collected the data. MMH and JEC analyzed the data and wrote the paper. All authors participated in the design and conduct of the experiment and in the data interpretation. None of the authors had personal or financial conflicts of interest.
REFERENCES