Protein-Enriched Liquid Preloads Varying in Macronutrient Content Modulate Appetite and Appetite-Regulating Hormones in Healthy Adults 1 3

Transcription

1 The Journal of Nutrition Ingestive Behavior and Neurosciences Protein-Enriched Liquid Preloads Varying in Macronutrient Content Modulate Appetite and Appetite-Regulating Hormones in Healthy Adults 1 3 Anestis Dougkas* and Elin Östman Food for Health Science Center, Lund University, Lund, Sweden Abstract Background: Dietary protein is considered the most satiating macronutrient, yet there is little evidence on whether the effects observed are attributable to the protein or to the concomitant manipulation of carbohydrates and fat. Objective: The aim was to examine the effect of consumption of preloads varying in macronutrient content on appetite, energy intake, and biomarkers of satiety. Methods: Using a randomized, within-subjects, 2-level factorial design, 36 adults [mean 6 SD age: y; body mass index (in kg/m 2 ): ) received a breakfast consisting of 1 of 7 isovolumetric (670 ml) and isoenergetic (2100 kj) liquid preloads matched for energy density and sensory properties but with different macronutrient composition (levels: 9%, 24%, or 40% of energy from protein combined with a carbohydrate-to-fat ratio of 0.4, 2, or 3.6, respectively). Appetite ratings and blood samples were collected and assessed at baseline and every 30 and 60 min, respectively, until a lunch test meal, which participants consumed ad libitum, was served 3.5 h after breakfast. Results: Prospective consumption was 12% lower after intake of the high-protein (40%)/3.6 carbohydrate:fat preload than after intake of the low-protein (9%)/0.4 carbohydrate:fat preload (P = 0.02) solely because of the increased protein, irrespective of the manipulation of the other macronutrients. Most appetite ratings tended to be suppressed (13%) with increasing protein content of the preloads (P < 0.06). Carbohydrate elicited greater increases in fullness and postprandial responses of glucose and insulin than did protein and fat. The glucose concentration was suppressed and glucagon-like peptide 1 increased more after intake of the high-protein (40%)/0.4 carbohydrate:fat preload than after the other preloads (P < 0.001). No statistically significant differences in postprandial ghrelin release or ad libitum energy intake at lunch were found. Conclusions: By varying all 3 macronutrients simultaneously and in a systematically balanced manner, we found that protein had a more pronounced effect on suppressing appetite than did carbohydrates and fat. Modulating the nutritional profile of a meal by replacing fat with protein can influence appetite in healthy adults. This trial was registered at www. clinicaltrials.gov as NCT J Nutr 2016;146: Keywords: Introduction appetite, macronutrients, GLP-1, protein, design of experiments Obesity has been increasing at an alarming rate, with >0.5 billion people being obese [BMI (in kg/m 2 ) $30] (1). Given that control of energy intake is vital to energy balance, understanding 1 Funded by Lund University Antidiabetic Food Center, a Swedish Innovation Agency (VINNOVA) Excellence Center. 2 Author disclosures: A Dougkas and E Östman, no conflicts of interest. 3 Supplemental Table 1, Supplemental Figures 1 3, and Supplemental Method are available from the Online Supporting Material link in the online posting of the article and from the same link in the online table of contents at jn.nutrition. org. *To whom correspondence should be addressed. anestis.dougkas@ food-health-science.lu.se. the appetite control system is important to combat the obesity epidemic. Within the past few years, there has been an increased interest in the macronutrient profile of diets and meals as an important factor that influences satiety (sensation of fullness) (2, 3). Several studies have suggested that the satiating power of the macronutrients follows a hierarchy in which protein is more satiating than carbohydrates and carbohydrates more satiating than fat (2, 4, 5), yet others failed to show this effect (6 8). Possible reasons for discrepancies in study outcomes include heterogeneities in study design (e.g., different preload time intervals, sources of macronutrients, energy density, volume, and palatability) and subjectsõ characteristics (e.g., sex and BMI) (9). ã 2016 American Society for Nutrition. Manuscript received May 18, Initial review completed August 13, Revision accepted December 11, First published online January 20, 2016; doi: /jn

2 These differences make it difficult to pinpoint the optimum dose or energy percentage needed to observe significant effects of any of the macronutrients on satiety. Evidence suggests that high doses of protein ($50 g in a meal) are necessary to see a significant effect on satiety, but there is currently insufficient information to describe a dose-response relation (5, 10, 11). Furthermore, it still remains elusive how variation in the quantity of macronutrients affects the satiety-implicated mechanisms. One putative mechanism is the release of gastrointestinal hormones, which are known to reduce gut motility, gastric emptying, and appetite (12). Such hormones include the appetite-suppressing glucagon-like peptide 1 (GLP-1) 4, peptide tyrosin-tyrosin, and the appetite-stimulating hormone ghrelin (13). There is evidence that protein and carbohydrate ingestion suppresses postprandial ghrelin response more than fat (14, 15) and that protein and fat induce higher release of peptide tyrosintyrosin than carbohydrates do (16). However, there is no consensus on which macronutrient is the greater stimulator of postprandial GLP-1 release (17, 18). Most studies examining the effects of macronutrients on satiety varied the content of 2 macronutrients, keeping the energy from the third macronutrient fixed. This type of study design makes it difficult to draw conclusions on whether the effects observed are attributable to the change in, for example, protein or to the concomitant manipulation of carbohydrates and/or fat (19). The independent effect of a macronutrient on satiety needs to be considered in a dose-response manner by comparing all 3 macronutrients simultaneously. Therefore, the main objective of the current study was to investigate the impact of preloads varying in all 3 macronutrients in a systematically balanced manner on measures of appetite and ad libitum energy intake. In addition, some potential mechanisms underlying any observed effects were examined. Methods Subjects. Forty healthy, nonsmoking subjects (23 men and 17 women, BMI , age y) were recruited through advertisements on notice boards and social media at Lund University (Supplemental Figure 1). Subjects did not have any known form of cardiovascular disease or metabolic or gastrointestinal disorder and were not taking any medications that could affect appetite or weight regulation. They were selected to be nonvegetarian nondieters (weight change of <3 kg during the past 2 mo) and noncognitively restrained eaters as defined by a score of #12 on factor 1 of the Three-Factor Eating Questionnaire (20) (Supplemental Method). The study was conducted according to the guidelines in the Declaration of Helsinki, and all procedures were approved by the Regional Ethical Review Board in Lund, Sweden (reference number 2012/6). Written informed consent was obtained from all subjects before commencing the study. All subjects were remunerated [3750 SEK ($440), of which 520 SEK ($61) were as money off food coupons] for their participation in the study. This trial was registered at gov as NCT Study design. This was a single-blind, randomized by a study-independent statistician (random permuted balanced blocks; size of 7), withinsubject study. All subjects attended the Food for Health Science Center on 7 separate occasions at least 1 wk apart. Subjects were instructed to maintain their normal dietary habits and physical activity level throughout the study period. There were no restrictions other than 4 Abbreviations used: DoE, design of experiments; GLP-1, glucagon-like peptide 1; HC, high carbohydrate; HF, high fat; HP, high protein; LC, low carbohydrate; LF, low fat; LP, low protein; LSM, least squares mean; MC, medium carbohydrate; MF, medium fat; MP, medium protein; VAS, visual analog scale. 638 Dougkas and Östman keeping a similar diet and refraining from alcohol intake and intensive exercise for 24 h before each study visit. A standardized ready meal [ham and cheese pizza (Orkla Food), 5.1 MJ; 18% energy from protein, 44% energy from carbohydrates, and 40% energy from fat) was supplied by the study, and subjects were asked to consume as much pizza as they felt comfortably satisfied on the evening before the study visits. After a 12-h overnight fast, subjects arrived at the department at 0800 and completed a 24-h food, physical activity, sleep, and mood questionnaire to assess compliance with the instructions and any potential factors that could influence food intake. If anything atypical was observed, the study visit was rescheduled for the following week. Female subjects were asked to complete a menstrual cycle questionnaire and report the first date of their last menstrual period at each study visit. Anthropometric [weight (kg), height (cm), and waist circumference (cm)] and body composition measurements using the bioimpedance method (BC-418MA; TANITA) were performed, followed by insertion of an indwelling cannula into the antecubital vein of the forearm. After a 20-min rest, blood pressure was measured 3 times, and the mean was used (Omron M6 Comfort; Omron Healthcare). A baseline venous and capillary blood sample, respectively, was collected and appetite, mood, and well-being were assessed with a computerized visual analog scale (VAS). One of the 7 test liquid preloads was then served chilled in opaque cups with a lid and straw to minimize visual comparisons. Timing started when the preload was provided to the subjects (0 min), and they were instructed to consume it within 15 min. Appetite and mood were assessed at 30, 60, 90, 120, 150, 180, and 210 min; well-being was assessed at 120 and 210 min; and blood samples were taken at 30, 90, 150, and 210 min from preload initiation. Subjects were allowed to read, do desktop work, and listen to music while they remained seated, but they were not allowed to drink or eat anything during the experimental session other than a glass of noncarbonated water (200 ml), which was provided at 115 min at each study session. Energy intake was assessed by an ad libitum pasta test meal offered in moderate excess (4 portions of standard serving size) to the subjects 210 min after preload initiation. Subjects ate in individual dining booths without any distractions and were asked to stop eating when they felt comfortably satisfied. The time to consume the test meal was 25 min. The palatability and sensorial properties of the preloads and test meal were assessed immediately after their consumption. The time interval between the preloads and the test meal was selected based on previous research (2, 21). Test preloads and meal. To investigate the effect of macronutrients on appetite and energy intake, we constructed a carefully prepared set of representative experiments [design of experiments (DoE)], in which all relevant factors were varied simultaneously. The first factor was the percentage of energy from protein with 2 levels [9% low protein (LP) and 40% high protein (HP)], and the second factor was the carbohydrateto-fat ratio (carbohydrate:fat) with 2 levels: 1) 0.44, low carbohydrate (LC) to high fat (HF) or LC to medium fat (MF) and 2) 3.56, high carbohydrate (HC) to low fat (LF) or medium carbohydrate (MC) to LF. A 2-level full factorial design was carried out and resulted in 7 experiments and the development of 5 test preloads, 4 points of the factorial design (LP/LC:HF, HP/LC:MF, LP/HC:LF, and HP/MC:LF), and 1 center point [medium protein (MP)/MC:MF]. The center point, consisting of 24% energy from protein (MP) and 2.0 carbohydrate:fat (MC:MF), was performed in triplicate [(1 3) indicating the 3 identical MP/MC:MF test preloads] to establish the experimental error and thus make possible an estimation of reproducibility. The range of the macronutrients was selected based on the experimental feasibility to minimize the differences in the texture and organoleptic properties among the treatments. Furthermore, an effort was made to make the investigation range large enough to allow the capturing of the effect of each factor. Using this balanced and orthogonal arrangement of experiments enabled the effect of one factor to be assessed independently of the other factors and the estimation of the interaction between factors. In addition, using the triplicates of the center point, the results were evaluated in the light of the systematic variability (noise). According to the Swedish Nutrition Recommendations and the average range of energy that a breakfast meal provides (22), the test preloads were equivalent to 20% of the Recommended Daily Intake for

3 men (2093 kj) and women (1675 kj), respectively. All the preloads were isoenergetic and isovolumetric, and they had the same energy density. The composition of the macronutrient-manipulated preloads is given in Table 1. Milk protein isolate (Glanbia Nutritionals), rapeseed oil (AAK), and a combination of maltodextrin (C*DryLight MD 01970; Cargill Nordic A/S) and regular white sugar (ICA) were used. No dietary fibers were included. The sensory properties were optimized using artificial sweetener (sucralose), salt, vanillin, and peach flavor. In addition, for further taste enhancement, all ingredients were dissolved in a mixture of 9.4 parts of tap water to 0.6 parts of tropical fruit flavored light squash drink (Fun Light; Orkla Food). The flavor, texture, and sensorial properties of the preloads were matched as closely as possible and tested in a pilot study using a panel of 20 people working at Orkla Foods in Eslöv, Sweden. There were no differences in pleasantness and overall sensory perception among the test preloads (data not shown). All the test preloads were prepared fresh the day before each study day and stored in a refrigerator (4 C) until served. The lunch test meal, which participants consumed ad libitum, was a single course, which was provided in excess (7189 kj) and composed of 500 g fresh cheese and ham tortellini (ICA), tomato and basil sauce (Dolmio; Mars Food UK Ltd.), and grated cheese (ICA). The energy content of the test meal was 1797 kj/338 g of serving portion, and macronutrient composition was 17 g protein (16% of energy), 62 g carbohydrate (57% of energy), and 13 g fat (27% of energy). VAS. Subjective appetite profile was assessed using 100-mm VAS ratings of hunger, desire to eat, fullness, and prospective food consumption anchored by the terms not at all and extremely. A composite appetite score was also calculated at each time point to reflect the 4 questions using the following formula: average appetite (mm) = [hunger + desire to eat + (100 2 fullness) + prospective consumption]/4 (23). Sensory and hedonic VAS ratings were collected in terms of pleasantness, enjoyment, visual appeal, taste, smell, difficulty in consumption, and overall palatability immediately after the preloads and test meal ingestion. Furthermore, recommended and validated questionnaires of mood and gastrointestinal well-being VAS ratings (Supplemental Method) were recorded (24, 25) and performed electronically on laptops using the Adaptive Visual Analogue Scales (Neurobehavioral Research Laboratory Clinic, USA) software (26) to increase data capture reliability. All subjects were trained on how to use the software before the study commenced. Blood sample collection and analysis. At screening and study days, capillary blood samples were taken by finger prick and immediately analyzed to quantify glucose concentrations (Glucose 201+; Hemocue AB). During the study days, venous blood was collected into chilled K 2 TABLE 1 Nutritional composition of the test preloads for men and women 1 EDTA tubes containing Dipeptidyl Peptidase 4 (DPP-IV) inhibitor (Merck Millipore) and Pefabloc (Roche Diagnostics Scandinavia) to prevent gut hormone degradation and proteases formation. The blood tubes were centrifuged at g for 15 min at 4 C, and plasma was separated and stored at 280 C until analysis. Plasma total ghrelin, insulin, and total GLP-1 concentrations were determined using the Bio- Plex human diabetes magnetic bead assay (Bio-Rad Laboratories) and analyzed on a Luminex 200 system (Luminex Corporation). Intra- and interassay CVs for insulin were 4.7% and 10%, respectively, for GLP-1 were #10% and #15%, respectively, and for ghrelin were #10% and #20%, respectively. Statistical analyses. Taking a 2-sided 5% significance level and having 80% probability, power analysis indicated that a total of 35 subjects should enter this 7-treatment within-subject study. The power was calculated to detect a difference of 10 mm in appetite ratings and 500 kj in energy intake between treatments, with a within-subject variance of 20 mm and 950 kj for appetite ratings and energy intake, respectively (27, 28). The DoE procedure was created and conducted using the statistical software MODDE (version 10.0; Umetrics AB). A full factorial design was developed to evaluate which factors (protein, carbohydrate:fat) were more important as well as their optimal ranges. The relation between the macronutrient level or their interactions and the responses (appetite variables or biochemical markers) were analyzed using multiple linear regression. Coefficients and response contour plots were used to interpret the influence of the factors. The effects of the 7 test preloads on energy intake and subjective appetite, mood, and well-being ratings were analyzed by using a linear mixed-model ANCOVA (PROC MIXED procedure). The fixed effects included in the model were baseline (pretreatment), treatment, visit, time, and treatment 3 time interaction. Subject was treated as a random effect, whereas time and visit were included in the repeated effects. The significant effect of sex, sex 3 treatment, sex 3 visit, body weight, and the stage of menstrual cycle was tested and included as a covariate when found significant (model 1). Further adjustment for the mood, appearance, and palatability variables was made and model 2 derived after backward stepwise elimination of those variables and/or interactions, which were not significant (approximate F tests). Pearson correlations in all mood, hedonic, and sensory variables were performed before backward stepwise analysis to determine and avoid multicollinearity. Only those variables that were not highly correlated (r < 0.7) were included in the analysis. All models were tested for the normality of residuals, and standard diagnostics were used to ensure that all variables met the normal distribution assumption. AUCs for appetite sensations and biomarkers of satiety were calculated for each subject and test LP/LC:HF 1 HP/LC:MF LP/HC:LF HP/MC:LF MP/MC:MF (1 3) 2 Men Women Men Women Men Women Men Women Men Women Energy, kj Weight, g Energy density, kj/g Protein, g Carbohydrate, g Maltodextrin, g Sugar, g Fat, g Protein, % energy Carbohydrate, % energy Fat, % energy Carbohydrate-to-fat ratio HC, high carbohydrate; HF, high fat; HP, high protein; LC, low carbohydrate; LF, low fat; LP, low protein; MC, medium carbohydrate; MF, medium fat; MP, medium protein. 2 (1 3) indicate the 3 identical MP/MC:MF test preloads representing the center point of the experimental design. The triplicates included 24% of energy from protein (MP) and a 2.0 carbohydrate-to-fat ratio (MC:MF). Differential effects of macronutrients on appetite 639

4 preload by using the trapezoid model. The relation between macronutrients and AUCs of appetite sensations and appetite-regulating hormones was examined by using a repeated-measures linear mixed model and was described by using fixed-effects coefficients, with the high percentage of energy from protein and carbohydrate:fat, respectively, being the reference level for the 2 factors. Tukey-Kramer post hoc significance test was performed to adjust for multiple comparisons of significant effects, and the Kenward-Roger correction was applied for reducing small sample bias. Data are presented as least squares means (LSMs) and SEMs, unless otherwise specified. Statistical significance was considered at P < 0.05 (2-tailed). The statistical package SAS (version 9.2; SAS Institute, Inc.) was used for all calculations. Results Subjects and baseline characteristics. Of the 40 subjects participating in the study, 36 completed the 7 preload conditions and were included in the analysis data set. The 4 withdrawals were due to personal reasons (n = 3) or protocol nonadherence (n = 1). The baseline characteristics of the study participants are shown in Table 2. Hedonic evaluation of the preloads. No differences were found in any of the hedonic and sensorial characteristics between the preloads (P $ 0.05) (Table 3), and the results showed that the scores were moderate (>50 mm). Furthermore, there were no differences between the preloads in all gastrointestinal well-being ratings throughout the study day (P $ 0.05; data not shown). Subjective appetite profile. Baseline ratings for none of the appetite variables were different between the treatments (P $ 0.05). The subjective ratings of the average appetite (Supplemental Figure 2) and the means of hunger, desire to eat, fullness, and prospective food consumption over the morning ( min) are shown in Table 4. In model 1, adjustment for baseline values, visit, sex, treatment, time, and treatment 3 time interaction was made. There was no significant effect of preload on hunger (P = 0.11) and a borderline significant effect on desire to eat (P = 0.06) and overall appetite (P = 0.06). Fullness ratings were 18% (P = 0.030) higher after the intake of HP/ MC:LF than after the intake of LP/LC:HF. Similarly, prospective food consumption was 12% and 11% (P < 0.05) lower after the intake of HP/MC:LF and HP/LC:MF relative to LP/LC:HF, respectively. After elimination of all the nonsignificant mood and palatability variables (model 2), the average appetite reached significance (13% decrease after intake of HP/MC:LF than LP/LC:HF, P < 0.05). All subjective TABLE 2 Baseline characteristics of the study participants 1 Subject characteristics Men (n = 21) Women (n = 15) Age, y Body weight, kg BMI, kg/m Waist, cm Fat mass, % Systolic blood pressure, mm Hg Diastolic blood pressure, mm Hg Capillary blood glucose, mmol/l TFEQ F1, cognitive restraint Values are means 6 SDs. F1, Factor 1, cognitive restraint; TFEQ, Three-Factor Eating Questionnaire. 640 Dougkas and Östman appetite ratings displayed a significant main effect of time in response to the treatment but no treatment 3 time interaction (data not shown). Ad libitum energy intake. No evidence for a difference in the mean ad libitum energy intake was found between the test preloads (P = 0.65). The overall mean energy intake (weight) at lunch 210 min after preload was kj ( g). There was a sex effect, with the mean energy intake over the 7 study days for men and women being kj ( g) and kj ( g), respectively. Glycemic and hormonal responses. No differences in baseline concentrations of the biomarkers of satiety were found between the preloads. There was a significant treatment effect (P < 0.001) and a time 3 treatment interaction (P < 0.001) for blood glucose and insulin responses (Figure 1). Comparisons of adjusted LSMs between the preloads showed that glucose response was lower after consumption of HP/LC:MF and increased to a greater degree after the consumption of LP/HC: LF than after the other preloads (Table 5). Insulin response was higher after HP/MC:LF and lower after LP/LC:HF than the other preloads. Specifically, insulin concentration was 43% (P < 0.001) and 24% (P < 0.001) lower after LP/LC:HF and HP/LC: MF than HP/MC:LF, respectively. A significant treatment effect (P < 0.001) and time 3 treatment interaction (P < 0.001) were found for total GLP-1 but not for total ghrelin (treatment effect, P = 0.35; time 3 treatment, P = 0.45) (Figure 1). Comparisons of adjusted LSMs between the preloads showed that the GLP- 1 concentration was higher and remained more elevated after HP/LC:MF than after the other preloads, with differences in GLP-1 concentration between HP/LC:MF and the other preloads ranging between 13% and 19% (P < 0.05) (Table 5). Furthermore, the GLP-1 concentration was 11% higher after HP/LC:MF relative to HP/MC:LF, but this difference was of borderline significance (P = 0.07). More pronounced differences in GLP-1 concentration between the HP/LC:MF and the LP/HC: LF or 2 of the center points were found at 150 min (Figure 1C). Although ghrelin was suppressed and peak concentration was reached at 90 min after intake of LP/LC:HF, HP/LC:MF, LP/HC: LF, and MP/MC:MF1 2 and at 150 min after intake of HP/MC: LF and MP/MC:MF3 relative to the other time points (Figure 1D), there were no differences in ghrelin between preloads throughout the morning. There was neither a sex effect (P > 0.05) nor differences in glucose and blood hormones between the 3 identical test preloads MP/MC:MF1 3 representing the center point. Contribution of macronutrients to appetite, appetiteregulating hormones, and glucose. The differences of the LSMs of the AUC (0 210 min) for all appetite sensations between low (9%) and high (40%) energy from protein and between low (0.4) and high (3.6) carbohydrate:fat are shown in Table 6. Hunger increased most when both the protein and carbohydrate:fat were low. However, given that only the change in protein was significant (P = 0.039), the increase in hunger could be attributable to the decreased protein content. Increased carbohydrates contributed to the increased sensation of fullness, given that fullness was lower after consumption of the 3.6 carbohydrate:fat as indicated by the negative estimate and the significant 2-fold difference in carbohydrate:fat (P = 0.007) compared with protein. On the contrary, prospective consumption was significantly suppressed because of the increase in protein content ( mm 3 min lower after intake of

5 TABLE 3 Hedonic ratings from visual analog scales after intake of LP/LC:HF, HP/LC:MF, LP/HC:LF, HP/MC:LF, or MP/MC:MF (1 3) preloads as breakfast by healthy adult men and women 1 Treatment LP/LC:HF HP/LC:MF LP/HC:LF HP/MC:LF MP/MC:MF (1) MP/MC:MF (2) MP/MC:MF (3) P value Pleasantness, mm Enjoyment, mm Difficulty in consumption, 2 mm Taste, mm Overall palatability, mm Aftertaste, mm Values are least squares means 6 SEMs, n = 36. (1 3) indicate the 3 identical MP/MC:MF test preloads representing the center point of the experimental design. The triplicates included 24% of energy from protein (MP) and a 2.0 carbohydrate-to-fat ratio (MC:MF). HC, high carbohydrate; HF, high fat; HP, high protein; LC, low carbohydrate; LF, low fat; LP, low protein; MC, medium carbohydrate; MF, medium fat; MP, medium protein. 2 How hard was it to consume the given amount of preload. 9% compared with 40% energy from protein, P =0.001) independent of the increase in carbohydrate content (P = 0.60). Postprandial glucose response was suppressed when the percentage of energy from protein increased and the carbohydrate:fat decreased (P < 0.001). Furthermore, low percentage of energy from protein and carbohydrate:fat suppressed the postprandial insulin response (P < 0.001), with the change in carbohydrate:fat having a greater effect on both glucose and insulin response than the change in percentage of energy from protein (Table 6). Postprandial GLP-1 response decreased at 9% of energy from protein and increased at 0.4 carbohydrate:fat, yet the bigger size effect of the difference in the percentage of energy from protein (211, pg 3 min/ml, P = 0.001) compared with the corresponding value of carbohydrate:fat ( pg 3 min/ml, P = 0.026) indicated the greatest influence of protein content. DoE for appetite, appetite hormones, and glucose responses. The 2 factors of interest (percentage of energy from protein and carbohydrate:fat) were varied simultaneously and independently using the DoE to investigate how the macronutrients influence appetite responses and appetite-regulating hormones and to identify their optimal ranges (Supplemental Table 1). Most models had good predictive ability and are considered reliable based on the R 2 and Q 2 (preferably high R 2 and Q 2 > 0.5 and difference R 2 2 Q 2 < 0.3) except for the AUC models of fullness, ghrelin, and GLP-1. Regarding the subjective appetite sensations, AUCs of prospective consumption and appetite were inversely influenced by protein (P < 0.01), and given the higher absolute value of the coefficient, protein had the greatest effect, particularly on reducing prospective consumption. Although carbohydrate:fat tended to also decrease the AUC of the appetite response, that decrease was not significant (P = 0.08). TABLE 4 Subjective appetite responses from visual analog scales after intake of LP/LC:HF, HP/LC:MF, LP/HC:LF, HP/MC:LF, or MP/ MC:MF (1 3) preloads as breakfast ( min) by healthy adult men and women 1 Treatment LP/LC:HF HP/LC:MF LP/HC:LF HP/MC:LF MP/MC:MF (1) MP/MC:MF (2) MP/MC:MF (3) Hunger, mm Model Model Desire to eat, mm Model Model Fullness, mm Model b a,b a,b a a,b a,b a,b Model b a,b a,b a a,b a,b a,b Prospective consumption, mm Model a b a,b b a,b a,b a,b Model a b a,b b a,b a,b a,b Appetite, mm Model Model a a,b a,b b a,b a,b a,b P value 1 Values are least squares means 6 SEMs, n = 36. (1 3) indicate the 3 identical MP/MC:MF test preloads representing the center point of the experimental design. The triplicates included 24% of energy from protein (MP) and a 2.0 carbohydrate-to-fat ratio (MC:MF). Model 1 was adjusted for baseline values, visit, sex, treatment, time, and treatment 3 time interaction. Model 2 backward stepwise analysis by eliminating the nonsignificant variables of mood, appearance, and palatability. Labeled means within a row without a common superscript letter differ, P, 0.05 (ANCOVA followed by TukeyÕs post hoc test). HC, high carbohydrate; HF, high fat; HP, high protein; LC, low carbohydrate; LF, low fat; LP, low protein; MC, medium carbohydrate; MF, medium fat; MP, medium protein. Differential effects of macronutrients on appetite 641

6 FIGURE 1 Concentrations of blood glucose (A) and plasma insulin (B), GLP-1 (C), and ghrelin (D) throughout the study day after consumption of LP/LC:HF, HP/LC:MF, LP/HC:LF, HP/MC:LF, or MP/MC:MF (1 3) preloads as breakfast by healthy adults. (1), (2), and (3) indicate the 3 identical MP/MC:MF test preloads representing the center point at the experimental design. Values are means 6 SEMs, n = 36. Labeled means at a time without a common lowercase letter differ, P, 0.05 (ANCOVA followed by TukeyÕs post hoc test). GLP-1, glucagon-like peptide 1; HC, high carbohydrate; HF, high fat; HP, high protein; LC, low carbohydrate; LF, low fat; LP, low protein; MC, medium carbohydrate; MF, medium fat; MP, medium protein. Considering glucose and the appetite-relating hormones, the AUC model of glucose had the best predictive ability, and analysis of the coefficients indicated that both protein and carbohydrate:fat contributed equally to the decrease and increase of postprandial glucose, respectively (P < 0.01). Postprandial insulin was positively influenced by carbohydrate: fat, which was the main effect (P = 0.007), whereas GLP-1 (AUC min) was positively influenced by protein and negatively by carbohydrate:fat. However, the effect of protein on GLP- 1 did not reach significance (P = 0.06), which resulted in an overall poor model. Although this analysis was crude by using the sum of all available data, it provided further insight and confidence of the independent effect of the macronutrient manipulations obtained using the mixed model analysis. To gain a better understanding of the modeled system, response contour plots of the optimal regression models with the factors percentage of energy protein and carbohydrate:fat are displayed in Supplemental Figure 3. The response contour plot of prospective consumption revealed that a low AUC of prospective consumption (9348 mm 3 min) was achieved at a high percentage of energy from protein (>38%), independent of carbohydrate:fat. However, low appetite (9515 mm 3 min) was observed at a percentage of energy from protein of >37% and carbohydrate:fat >3.0. Discussion To our knowledge, this is the first study to use the DoE approach, in which all 3 macronutrients were varied simultaneously and in a systematically balanced manner. Consequently, it was possible to ascribe the effect of any of the macronutrients on appetite, energy intake, and biomarkers of satiety, independent of changes in the other macronutrients. The main findings were as follows: 1) prospective consumption was 12% lower after intake of the HP/MC:LF (40%/3.6 carbohydrate:fat) compared with the LP/LC:HF (9%/0.4 carbohydrate:fat) preload solely because of the increased protein, irrespective of the manipulation of the other macronutrients; 2) there was no difference in ad libitum energy intake after the different test preloads; and 3) carbohydrates appeared to be the most influential macronutrient for postprandial responses of glucose and insulin, whereas protein had the greatest influence on GLP-1. Although several studies examined the satiating capacity of the macronutrients with equivocal results (5 8, 29, 30), the present findings are in contrast to previous studies that have shown no differences in appetite sensations after consumption of protein, carbohydrate, or fat-enriched meals/preloads (6 8, 18, 31). Methodologic differences such as palatability and sensory properties of the preloads could act as potential sources that may partly explain the inconsistent findings (9). For instance, de Graaf et al. (6) showed similar hunger ratings and energy intake between the 3 macronutrients, yet the subjects were asked to use nose clips because of the substantial differences in the smell, taste, and palatability of the preloads. In addition to the fact that increased palatability has been shown to increase energy intake, bypassing the orosensory and olfactory cues using nose clips could also influence appetite (32). Although the palatability ratings of the preloads in the present study were moderate, careful study design and extensive product development aimed to avoid potential confounding factors by employing indistinguishable, isoenergetic, and isovolumetric preloads that were varied covertly for their macronutrient content but matched for energy density and sensory properties. 642 Dougkas and Östman

7 TABLE 5 Circulating concentrations of glucose and gastrointestinal hormones after intake of LP/LC:HF, HP/LC:MF, LP/HC:LF, HP/MC:LF, or MP/MC:MF (1 3) preloads as breakfast ( min) by healthy adult men and women 1 Treatment LP/LC:HF HP/LC:MF LP/HC:LF HP/MC:LF MP/MC:MF (1) MP/MC:MF (2) MP/MC:MF (3) P value Glucose, mmol/l c d a b,c b b b,0.001 Insulin, pmol/l c b a a a a,b a,b,0.001 GLP-1, pg/ml b a b a,b b b b,0.001 Ghrelin, pg/ml Values are least square means 6 SEMs. n = 36. (1 3) indicate the 3 identical MP/MC:MF test preloads representing the center point of the experimental design. The triplicates included 24% of energy from protein (MP) and a 2.0 carbohydrate-to-fat ratio (MC:MF). Labeled means within a row without a common superscript letter differ, P, 0.05 (ANCOVA followed by TukeyÕs post hoc test). GLP-1, glucagonlike peptide 1; HC, high carbohydrate; HF, high fat; HP, high protein; LC, low carbohydrate; LF, low fat; LP, low protein; MC, medium carbohydrate; MF, medium fat; MP, medium protein. Interestingly, protein appeared to be the most influential macronutrient for prospective consumption because both test preloads with high protein content (HP/LC:MF and HP/MC:LF) significantly suppressed prospective consumption compared with the LP/LC:HF. However, there were no differences in the other appetite sensations between the HP/LC:MF and the LP/ LC:HF. Moreover, there were no differences in appetite sensations between HP/MC:LF and LP/HC:LF, which indicates that carbohydrate content of the preload acts synergistically with protein for most of the appetite sensations. In addition to the studies that showed no differences between the macronutrients, studies showed that protein and carbohydrate have similar and higher satiating capacity than fat, assessed either by VAS scores (30, 33) or subsequent energy intake (8, 29, 34). However, the average appetite in the present study was mainly driven by the increased protein, rather than the increased carbohydrates and decreased fat content. Overall, the present findings clearly showed that independent of the modifications of protein or carbohydrate, fat is the least satiating macronutrient. This is in agreement with a few studies (30, 33) that showed increased hunger after intake of an HF meal relative to HP and HC meals of identical energy density. In contrast to those studies, we did not find differences between the macronutrients on energy intake at the subsequent lunch. However, although these studies used a solid test preload or meal, liquid preloads were employed in the current study. Although the literature on the impact of the physical form of the experimental meal on satiety is ambiguous (35, 36), a recent systematic review reported weak effects of liquid calories on satiety and energy compensation as a result of the different cognitive, hormonal, and gastrointestinal responses (37). Considering the dose-response effect of macronutrients on appetite sensations, the fact that there was a difference in prospective consumption and average appetite between the LP (9% energy) and the HP preload (40% energy) suggests that a higher amount of protein could be associated with a greater influence on appetite sensations, as has been seen in some (11, 38, 39) but not all previous studies (40, 41). Given that the present study provided a maximum amount of protein of 50 g or 40% energy from protein, the existence of a possible threshold effect of protein (10) above which no additional benefit of increased protein intake with respect to satiation is evident cannot be excluded. Differences in the metabolic and hormonal profiles might be responsible for the differential satiating effects of macronutrients during the postmeal interval (42, 43). In the present study, postprandial glucose and insulin profiles reflected the carbohydrate content, with increased carbohydrate resulting in the TABLE 6 Coefficients of the linear mixed models for the relations of percentage of energy from protein or CHO:fat with the AUCs (0 210 min) of subjective appetite profile and hormonal responses during the study day in healthy adult men and women 1 Coefficients Intercept P value 9% Protein 2 P value 40% Protein CHO:fat 4 P value 3.6 CHO:fat 5 Subjective appetite profile AUC hunger, mm 3 min , Reference Reference AUC desire to eat, mm 3 min , Reference Reference AUC fullness, mm 3 min , Reference Reference AUC prospective consumption, mm 3 min , Reference Reference AUC appetite, mm 3 min , Reference Reference Glycemic and hormonal responses AUC glucose, mmol 3 min/l , Reference ,0.001 Reference AUC insulin, pmol 3 min/l , ,0.001 Reference ,0.001 Reference AUC ghrelin, pg 3 min/ml 42, , Reference Reference AUC GLP-1, pg 3 min/ml 127, , , Reference Reference 1 Values are coefficients 6 SEs. n = 36. CHO:fat, carbohydrate-to-fat ratio; GLP-1, glucagon-like peptide 1. 2 The low percentage of energy from protein. 3 The high percentage of energy from protein. 4 The low CHO:fat preload. 5 The high CHO:fat preload. The high level of the two factors was used as the reference in the mixed model (ANCOVA followed by TukeyÕs post hoc test). Differential effects of macronutrients on appetite 643

8 greatest insulin and glucose responses. However, the HP/MC:LF induced the highest insulin response, whereas the HP/LC:MF induced one of the lowest insulin responses compared with the other preloads, highlighting the impact of carbohydrate content on the effect of protein on insulin. Although the results from studies that used HP diets and had constant carbohydrate content are equivocal (44, 45), using the novel design, we demonstrated that increased protein consumption stimulates insulin release. The synergistic effect of protein and carbohydrates on insulin release may depend on mechanisms related to either increased concentrations of amino acids (46) and/or increased sensitivity of pancreatic cells to glucose (47), possibly irrespective of the actions of incretin hormones (48, 49). The latter is further supported by the fact that the greatest and most sustained GLP-1 response was found after intake of the HP/LC: MF, which at the same time induced one of the lowest insulin responses compared with the other preloads. The current results showed that both increased protein and fat content increased postprandial GLP-1 response, with protein having the greatest effect, which is in agreement with some studies (11, 18, 50) but in contrast to others (51, 52). A hierarchy of macronutrientsõ ghrelin-suppressing capacity has been suggested, with carbohydrate > protein > fat (53). However, there is a notion that suppression of ghrelin after meals is influenced by its overall energy content (54, 55). The fact that the preloads were isoenergetic could thus explain the lack of differences in ghrelin response, which is in concordance with some studies (17, 18) but in contrast to others (2, 11, 39, 56). The differences in appetite responses observed after the intake of the various preloads appeared to be modest and were not followed by differences in energy intake at the subsequent lunch. Inclusion of both sexes induced significant variation in the ad libitum energy intake, which could mask any real effects. The test preloads were not typical breakfast foods, and it has been suggested that beliefs about the novelty and appropriateness of the foods influence satiety (8); thus, it may be a limitation of the study that cognitive cues and perceptions of the subjects regarding the type of preloads could also influence the results. The present findings cannot be generalized because there is evidence that within a group of macronutrients, the type and source of macronutrients may influence satiety, appetiteregulating hormones, and energy intake differently (9, 57, 58). In conclusion, examining the independent effects of macronutrient modifications on appetite sensations, protein appeared the most and fat the least influential macronutrient for prospective consumption and average appetite. Although there were no differences in subsequent energy intake, adjusting the nutritional profile of a meal, especially replacing fat with protein, could make dieting more endurable. Studies with complex designs using different quantities and sources or types of protein, carbohydrate, and fat and simultaneous measurements of their effects on appetite responses, energy intake, and appetiteregulating hormones would seem prudent. Acknowledgments We thank Olof Rosén (MKS Umetrics AB) and Susann Ullén (Region Skåne) for the valuable statistical help and advice, Marta Gisbert Guillem for assistance in the study, and Anders Högberg and Mats Krook (Orkla Foods) for their support in preload development. AD and EÖ contributed to the design of the study, and AD conducted the research and data analysis, drafted the manuscript, and had primary responsibility for final content. Both authors read and approved the final version of the manuscript. 644 Dougkas and Östman References 1. Ng M, Fleming T, Robinson M, Thomson B, Graetz N, Margono C, Mullany EC, Biryukov S, Abbafati C, Abera SF, et al. Global, regional, and national prevalence of overweight and obesity in children and adults during : a systematic analysis for the Global Burden of Disease Study Lancet 2014;384: Brennan IM, Luscombe-Marsh ND, Seimon RV, Otto B, Horowitz M, Wishart JM, Feinle-Bisset C. Effects of fat, protein, and carbohydrate and protein load on appetite, plasma cholecystokinin, peptide YY, and ghrelin, and energy intake in lean and obese men. Am J Physiol Gastrointest Liver Physiol 2012;303:G Li J, An R, Zhang Y, Li X, Wang S. Correlations of macronutrientinduced functional magnetic resonance imaging signal changes in human brain and gut hormone responses. Am J Clin Nutr 2012; 96: Paddon-Jones D, Westman E, Mattes RD, Wolfe RR, Astrup A, Westerterp-Plantenga M. Protein, weight management, and satiety. Am J Clin Nutr 2008;87:1558S 61S. 5. Poppitt SD, McCormack D, Buffenstein R. Short-term effects of macronutrient preloads on appetite and energy intake in lean women. Physiol Behav 1998;64: de Graaf C, Hulshof T, Weststrate JA, Jas P. Short-term effects of different amounts of protein, fats, and carbohydrates on satiety. Am J Clin Nutr 1992;55: Raben A, Agerholm-Larsen L, Flint A, Holst JJ, Astrup A. Meals with similar energy densities but rich in protein, fat, carbohydrate, or alcohol have different effects on energy expenditure and substrate metabolism but not on appetite and energy intake. Am J Clin Nutr 2003;77: Potier M, Fromentin G, Lesdema A, Benamouzig R, Tome D, Marsset- Baglieri A. The satiety effect of disguised liquid preloads administered acutely and differing only in their nutrient content tended to be weaker for lipids but did not differ between proteins and carbohydrates in human subjects. Br J Nutr 2010;104: Benelam B. Satiation, satiety and their effects on eating behaviour. Br Nutr Found Nutr Bull 2009;34: Anderson GH, Tecimer SN, Shah D, Zafar TA. Protein source, quantity, and time of consumption determine the effect of proteins on short-term food intake in young men. J Nutr 2004;134: Belza A, Ritz C, Sorensen MQ, Holst JJ, Rehfeld JF, Astrup A. Contribution of gastroenteropancreatic appetite hormones to proteininduced satiety. Am J Clin Nutr 2013;97: Cummings DE, Overduin J. Gastrointestinal regulation of food intake. J Clin Invest 2007;117: DelzenneN,BlundellJ,BrounsF,CunninghamK,DeGraafK,ErknerA, Lluch A, Mars M, Peters HP, Westerterp-Plantenga M. Gastrointestinal targets of appetite regulation in humans. Obes Rev 2010;11: Foster-Schubert KE, Overduin J, Prudom CE, Liu J, Callahan HS, Gaylinn BD, Thorner MO, Cummings DE. Acyl and total ghrelin are suppressed strongly by ingested proteins, weakly by lipids, and biphasically by carbohydrates. J Clin Endocrinol Metab 2008;93: Tannous dit El Khoury D, Obeid O, Azar ST, Hwalla N. Variations in postprandial ghrelin status following ingestion of high-carbohydrate, high-fat, and high-protein meals in males. Ann Nutr Metab 2006; 50: Cooper JA. Factors affecting circulating levels of peptide YY in humans: a comprehensive review. Nutr Res Rev 2014;27: Gibbons C, Caudwell P, Finlayson G, Webb DL, Hellstrom PM, Naslund E, Blundell JE. Comparison of Postprandial profiles of ghrelin, active GLP-1, and total PYY to meals varying in fat and carbohydrate and their association with hunger and the phases of satiety. J Clin Endocrinol Metab 2013;98:E van der Klaauw AA, Keogh JM, Henning E, Trowse VM, Dhillo WS, Ghatei MA, Farooqi IS. High protein intake stimulates postprandial GLP1 and PYY release. Obesity (Silver Spring) 2013;21: European Food Safety Authority (EFSA) Panel on Dietetic Products, Nutrition and Allergies. Scientific opinion on the substantiation of health claims related to protein and increase in satiety leading to a reduction in energy intake (ID 414, 616, 730), contribution to the maintenance or achievement of a normal body weight (ID 414, 616, 730), maintenance of normal bone (ID 416) and growth or maintenance of muscle mass (ID 415, 417, 593, 594, 595, 715) pursuant to Article 13 (1) of Regulation (EC) No 1924/2006. EFSA 2010;8:1811.