High consumption of sugar-sweetened beverages is associated with weight gain, obesity and increased risk of type 2 diabetes in both adults and children (Ludwig et al. Reference Ludwig, Peterson and Gortmaker2001; Bray et al. Reference Bray, Nielsen and Popkin2004; Schulze et al. Reference Schulze, Manson, Ludwig, Colditz, Stampfer, Willett and Hu2004). The energy content of carbonated sodas, in particular, does not appear to be compensated for during meals (Ludwig et al. Reference Ludwig, Peterson and Gortmaker2001). Some studies have suggested that liquids differ from solid foods in their effects on hunger, satiety and energy compensation (Almiron-Roig et al. Reference Almiron-Roig, Chen and Drewnowski2003). Soft drinks have thus been described as thirst-quenching liquids that somehow bypass the satiety mechanisms regulating food consumption (Almiron-Roig & Drewnowski, Reference Almiron-Roig and Drewnowski2003). However, the evidence is inconclusive. Other studies have suggested that energy compensation after an energy-containing beverage is instead dependent on the volume of the beverage and the timing of consumption before a meal (Almiron-Roig et al. Reference Almiron-Roig, Flores and Drewnowski2004). It has been argued, also, that some beverages have more of an effect on satiety than others. Hence sugar energy consumed in the form of fruit juice or milk may be more satiating compared with carbonated sugar-sweetened beverages (Rolls & Barnett, Reference Rolls and Barnett2000; Almiron-Roig & Drewnowski, Reference Almiron-Roig and Drewnowski2003).
The energy densities of low-fat (1 %) milk, orange juice and regular cola are almost identical (about 1·8 kJ/g); thus if these beverages differentially affect appetite and satiety then factors other than energy density must be involved. The present study aimed to compare the effect on appetite and energy intake of a sugar-sweetened beverage (cola) and a chocolate milk drink (containing protein and fat in addition to sugar), matched for energy content and volume. Studies with solid foods have indicated a satiety hierarchy for the macronutrients: protein>carbohydrate>fat (Rolls et al. Reference Rolls, Hetherington and Burley1988; Hill & Blundell, Reference Hill and Blundell1990; Hall et al. Reference Hall, Millward, Long and Morgan2003; Weigle et al. Reference Weigle, Breen, Matthys, Callahan, Meeuws, Burden and Purnell2005), but there is less evidence to support that this hierarchy holds true in liquids (Stubbs & Whybrow, Reference Stubbs, Whybrow, Wilson and Temple2003). In one study, consumption of a liquid meal replacement composed of sugar, fat and protein led to increased satiety than after an isoenergetic beverage containing only sugar (St-Onge et al. Reference St-Onge, Rubiano, Jones, Greenfield, Ferguson, Akrabawi and Heymsfield2004). Furthermore, hunger and desire to eat were less after a protein fruit juice beverage than after juices comprised of fat or carbohydrate (Westerterp-Plantenga & Verwegen, Reference Westerterp-Plantenga and Verwegen1999), although there was no difference in subsequent energy intake at lunch. However, another study found no difference between 1 % low-fat milk, cola and juice in terms of hunger and satiety ratings or energy intake at the subsequent lunch (Almiron-Roig & Drewnowski, Reference Almiron-Roig and Drewnowski2003), and no differences in appetite scores between beverages high in protein, carbohydrate or fat were reported in an earlier study (de Graaf et al. Reference de Graaf, Hulshof, Weststrate and Jas1992).
In the present cross-over randomised study of the effect on short-term appetite profile and subsequent energy intake of cola compared with chocolate milk, a preload-meal interval of 30 min was used. This interval has previously been associated with a compensatory effect on energy intake (Westerterp-Plantenga & Verwegen, Reference Westerterp-Plantenga and Verwegen1999). An ad libitum lunch, identical at each visit, was provided 30 min after consumption of the liquid preload. The hypothesis was that the chocolate milk would suppress appetite more than the cola because of its protein content.
Subjects and methods
Subjects
Male subjects were recruited for the study at the Panum Institute and The Royal Veterinary and Agricultural University (Copenhagen, Denmark) by means of advertisements. The inclusion criteria were: age 20–40 years, of normal weight (BMI range 18·5–25 kg/m2), non-smokers, not elite sportsmen, no previous history of diabetes, overweight, hypertension or liver disease, no daily use of medicine or excessive alcohol use. During recruitment, subjects were informed of the meals and preload beverages to be provided and asked to state any dislikes (none were noted). Twenty-two young men (age 23 (sd 1·8) years) of normal weight (BMI 22·2 (sd 1·5) kg/m2) were enrolled in the study and there was no drop-out. The study was approved by the Municipal Ethical Committee of Copenhagen and Frederiksberg and was in accordance with the latest version of the Helsinki Declaration. All subjects gave their written consent after the experimental procedure had been explained to them verbally and in writing.
Design
The study had a cross-over randomised design. Subjects were randomly assigned to whether the preload on the first visit was cola or chocolate milk (half in each group). On each of two test days, with at least 1 week between them, subjects reported at the Department of Human Nutrition in a fasted state. Strenuous physical activity was not permitted the day before each visit and it was requested that the evening meal be no later than 21.00 hours; water was permitted (up to 300 ml and 200 ml on the evening before and the morning of the study, respectively). On arrival at 08.00 hours, subjects were weighed and their height was measured. Subjects were then provided with breakfast, which consisted of porridge oats (Quaker Oats Ltd, c/o Consiva Foods A/S, Herlev, Denmark) with semi-skimmed milk, orange juice and either caffeine-free coffee, tea or water. An identical breakfast was provided on the second visit. Visual analogue scales (VAS) were used to record subjective appetite ratings every 30 min from 08.30 hours (t = 0), just before the breakfast meal, until 30 min after lunch (t = 240 min). The preload beverage of 500 ml cola or chocolate milk (matched for energy content at 900 kJ) was given at 11.30 hours (t = 180 min). VAS were also recorded just after intake of the beverage (t = 190 min) and again just before the ad libitum lunch (t = 210 min), which was provided at 12.00 hours. Ratings of hunger, satiety, fullness, prospective food consumption, thirst, desire for something salty, sweet or fatty, or some meat or fish were recorded on a 100 mm scale anchored with ‘not at all’ and ‘extremely’. How comfortable a subject felt was also recorded. The time interval of 30 min between preload and lunch was chosen based on a previous study, which showed that this interval had an effect with a 1·3 MJ preload (a milk shake compared with water or no preload) and that this effect was not due to volume (Westerterp-Plantenga & Verwegen, Reference Westerterp-Plantenga and Verwegen1999). Information about the appearance and palatability of the meals and preload beverages was also recorded within 10 min of ingestion.
Power analysis indicated that a sample of eighteen subjects should be sufficient in a cross-over study to detect a difference of 5 mm on mean 4.5 h VAS ratings, with a power of 80 % (Flint et al. Reference Flint, Raben, Blundell and Astrup2000). With twenty-two subjects, statistical significance of 0·05 and a power of 80 %, a difference in 660 kJ energy intake should be detectable at lunch 30 min after preload (Rolls et al. Reference Rolls, Kim, McNelis, Fischman, Foltin and Moran1991), with the reservation that this power analysis was based on comparison between a carbohydrate preload compared with no preload.
Test meals and beverages
The breakfast had an energy content of 2·5 MJ. Lunch, which was provided separately to each subject and which they could consume ad libitum until pleasantly full, was a pasta salad consisting of a homogeneous mix of pasta, ham, carrots, peas and dressing, prepared by the kitchen staff in the Department. The amount of food ingested at lunch was measured using a digital scale. Subjects had 20 min to eat both breakfast and lunch and the macronutrient composition of the meals was 15 % energy from protein, 30 % from fat and 55 % from carbohydrate. Intake of ad libitum water was allowed with lunch on the first visit and the amount for each subject was recorded; the same amount of water was provided on the second visit. The beverages consisted of Coca Cola (Coca Cola Nordic Services A/S, Hellerup, Denmark) and Matilde skimmed-milk chocolate milk (Arla Foods, Viby J, Denmark), both served cold. The energy content and nutrient composition of each preload beverage are shown in Table 1. The chocolate milk was presented with a separate glass of water so that the total energy content and volume was identical to that of the cola (chocolate milk of matching energy density was not available). Subjects were instructed to drink equally from the chocolate milk and the water glasses. The beverages in total had the same energy density (1·6 kJ/g) and supplied 900 kJ energy each. The cola contained energy only in the form of carbohydrate, whereas the chocolate milk also contained protein and fat. The study was designed so that there was a maximum of five subjects at the Department on a given test day, and subjects were kept apart, though in the same room, throughout the morning.
Table 1 Energy and macronutrient composition of the preload beverages
* Provided as 346 ml chocolate milk and 154 ml water in a separate glass.
Statistical analyses
The postprandial response curves were compared by ANOVA using mixed linear models with repeated measures. Post hoc ANOVA was used to test differences between the two groups (chocolate milk and cola) at individual time points; the factors were treatment, order and treatment × order. The mixed procedure in the Statistical Analysis System software package (version 9.1; SAS Institute, Cary, NC, USA) was used to test differences in appetite factors and energy intake. A significance level of P < 0·05 was used for all statistical tests (two-sided).
Results
Evaluation of test meals and beverages
There were no differences in subjective evaluation of the two preload beverages; both were equally well liked (in terms of appearance, aroma, taste and palatability) by the subjects. Overall palatability was 23 (sd 15) and 30 (sd 23) mm for chocolate and cola, respectively (0 = high palatability, 100 mm = low). Scores for the other parameters were similar. Similarly, scores for appearance and taste, etc, of the breakfast and lunch meals did not differ.
Subjective appetite ratings
Satiety and fullness, rated by subjective VAS scores, were significantly greater (F(1, 20) = 16·18, P = 0·0007 and F(1, 20) = 18·20, P = 0·0004, respectively) 30 min after the chocolate milk (just before lunch, t = 210 min) than after the cola drink (Fig. 1). Ratings of prospective consumption and hunger were significantly greater after cola than after chocolate milk, both just after intake of the beverage at t = 190 min (F(1, 20) = 8·69, P = 0·008, and F(1, 20) = 7·89, P = 0·01, respectively), and also just before lunch at t = 210 min (F(1, 20) = 10·82, P = 0·004, and F(1, 20) = 7·96, P = 0·01, respectively). The preference for savoury or high-protein foods was significantly lower after chocolate milk than after cola at t = 210 min (data not shown) (salty food: F(1, 20) = 9·24, P = 0·007; fatty food: F(1, 20) = 14·79, P = 0·001; meat or fish: F(1, 20) = 7·96, P = 0·01). No significant differences were observed in subjective appetite ratings at other time points of the study (P>0·05) and no differences in thirst, desire to eat something sweet or wellbeing were observed throughout the study.
Fig. 1 Subjective appetite sensations in twenty-two normal-weight men after intake of a preload beverage, chocolate milk (■) or cola (□), 30 min before an ad libitum lunch. A visual analogue scale rating equal to 100 mm corresponds to (A) ‘I cannot eat another bite’ (satiety); (B) ‘I have never been hungrier’ (hunger); (C) ‘I am totally full’ (fullness); (D) ‘I can eat a lot’ (prospective food consumption). Data are mean values, with their standard errors represented by vertical bars. By ANOVA, there was a treatment effect just after preload beverage (t = 190 min) for hunger (P = 0·02) and prospective consumption (P = 0·005). There was a treatment effect immediately before ad libitum lunch intake (t = 210 min) for satiety (P = 0·0007), hunger (P = 0·01), fullness (P = 0·0004) and prospective consumption (P = 0·004). There was no effect of time or treatment × time (P>0·05) for all parameters.
Ad libitum energy intake
Despite significant differences in subjective appetite ratings between preload beverages, these were not translated into differences in energy intake at lunch. Although there was overall a 4 % greater intake of ad libitum lunch after ingestion of cola (3286 (sd 1346) kJ) compared with chocolate milk (3145 (sd 1268) kJ), this difference was not significant (F(1, 20) = 0·68, P = 0·42; Fig. 2). Results were variable, with eleven subjects ingesting more after the chocolate milk and eleven ingesting more after cola; standard deviations were thus very high. Statistical analysis showed that there was no effect of treatment order (chocolate milk or cola on the first visit day) on any appetite parameter or energy intake at lunch.
Fig. 2 Spontaneous energy intake at an ad libitum lunch 30 min after consumption of a preload beverage, chocolate milk or cola, in twenty-two normal-weight men. Mean energy intakes at lunch were 3145 (sd 1268) kJ after the chocolate milk and 3286 (sd 1346) kJ after cola. There was no significant difference in energy intake at lunch (P = 0·42).
Discussion
The increased short-term satiety observed after the consumption of chocolate milk compared with cola was in contrast to a previous study, which found no differences in subjective appetite profiles between low-fat (1 %) milk, cola and orange juice (Almiron-Roig & Drewnowski, Reference Almiron-Roig and Drewnowski2003). In that study, subjects were given a slice of toast with the beverage, which provided 418 kJ out of the total 1454 kJ, and this could have masked the effect on appetite of beverage alone. Beverages in both studies were matched for energy density and energy content, but differed in their nutrient composition and sensory characteristics. Nutritional and sensorial differences between beverages could thus account for possible differential effects on satiety.
The protein content of the chocolate milk provides one explanation for the increased satiation we observed compared with sugar-rich cola, and supports the concept of a satiety hierarchy for the macronutrients: protein>carbohydrate>fat (Rolls et al. Reference Rolls, Hetherington and Burley1988; Hill & Blundell, Reference Hill and Blundell1990; Hall et al. Reference Hall, Millward, Long and Morgan2003; Weigle et al. Reference Weigle, Breen, Matthys, Callahan, Meeuws, Burden and Purnell2005). In a recent study, greater ratings of satiety were observed after consumption of a liquid formula composed of sugar, fat and protein than after an isoenergetic beverage containing only sugar (St-Onge et al. Reference St-Onge, Rubiano, Jones, Greenfield, Ferguson, Akrabawi and Heymsfield2004), and hunger and desire to eat were reduced after a protein fruit juice compared with after juices comprised of fat or carbohydrate (Westerterp-Plantenga & Verwegen, Reference Westerterp-Plantenga and Verwegen1999). In contrast, in another study no differences in appetite profiles were reported after beverages high in protein, carbohydrate or fat (de Graaf et al. Reference de Graaf, Hulshof, Weststrate and Jas1992), provided after an overnight fast. Differences in the timing of preload and the interval between preload and test meal could explain differences between studies (de Graaf et al. Reference de Graaf, Hulshof, Weststrate and Jas1992; Almiron-Roig et al. Reference Almiron-Roig, Flores and Drewnowski2004). Also, factors other than protein content could be involved.
One such factor could be glycaemic index (GI). Cola has a relatively higher GI than chocolate milk (about 58 v. 34) (Foster-Powell et al. Reference Foster-Powell, Holt and Brand-Miller2002) and some studies suggest that foods with a low GI are more satiating than foods with a high GI (Pawlak et al. Reference Pawlak, Ebbeling and Ludwig2002). Another factor is viscosity. A cross-over study in which subjects were given shakes matched for macronutrient content and energy density but which varied in viscosity (by addition of < 0·1 g methylcellulose) indicated that the thicker shake significantly reduced hunger compared with a thin shake (Mattes & Rothacker, Reference Mattes and Rothacker2001). Differences in sensory properties such as viscosity could also partly explain the greater prospective consumption and hunger ratings observed immediately after consumption of the cola in the present study (whilst satiety and fullness were the same for both beverages). The effect on appetite of the chocolate milk could also have been brought about by its content of cocoa powder, which could have elicited a raised insulin response. Chocolate products have been shown to elicit greater postprandial insulin secretion than products with alternate flavouring (Brand-Miller et al. Reference Brand-Miller, Holt, de Jong and Petocz2003). Chocolate milk elicited 45 % greater insulinaemia than strawberry flavoured milk, despite no difference in GI or glucose area under the curve. It was speculated that specific insulinogenic compounds in cocoa powder (as yet unknown) may directly stimulate β-cell insulin secretion, and it is interesting to note that high insulin responses have been associated with short-term appetite regulation and increased satiety in healthy subjects (Flint et al. Reference Flint, Gregersen, Gluud, Moeller, Raben, Tetens, Verdich and Astrup2005). Moreover, milk products in general have been shown to be insulinotropic (Ostman et al. Reference Ostman, Elmstahl and Bjorck2001). Some or all of the above factors could partly explain the satiating effect of chocolate milk in the present study.
Despite the fact that chocolate milk and cola differed in their effects on subjective appetite responses, there was no subsequent difference in energy intake at the ad libitum lunch. Similarly, increased short-term satiety after a protein fruit juice compared with juices comprised of fat or carbohydrate was not matched by reduced energy intake at lunch (Westerterp-Plantenga & Verwegen, Reference Westerterp-Plantenga and Verwegen1999). Studies which have demonstrated energy compensation after liquid preloads have used a short time interval between preload and test meal of up to 30 min (Kissileff et al. Reference Kissileff, Gruss, Thornton and Jordan1984; Rolls et al. Reference Rolls, Federoff, Guthrie and Laster1990, Reference Rolls, Kim, McNelis, Fischman, Foltin and Moran1991; Spiegel et al. Reference Spiegel, Hubert, Fried, Peikin, Siegel and Zeiger1997; Himaya & Louis-Sylvestra, Reference Himaya and Louis-Sylvestra1998), and volume also seems to be important (Rolls et al. Reference Rolls, Bell and Thorwart1998, Reference Rolls, Castellanos, Halford, Kilara, Panyam, Pelkman, Smith and Thorwart1999). Energy intake 30 min after a 1·3 MJ milk shake was significantly lower than after a water control or no preload, and this was assumed not to be due to volume (Westerterp-Plantenga & Verwegen, Reference Westerterp-Plantenga and Verwegen1999). The preloads in the present study had a lower energy content of 900 kJ, which could explain why we observed no energy compensation at lunch. Differences in sensory properties such as viscosity could also have had an effect. In addition, the present study was powered primarily to detect differences in VAS. It is possible that a difference in energy intake might have been detected with greater power. However, it cannot be ruled out that energy-containing beverages in general do not promote energy compensation at the next meal.
In the present study, a desire for salty or fatty food, or meat or fish, was significantly greater after cola than after chocolate milk. It is possible that had a buffet-style lunch been provided with more choice as to food components, then a difference in energy intake between the two beverages might have been observed. Provision of ad libitum meals in small studies such as this could be an inherent difficulty of the study design; changes in intake might in real life be more readily observed over time, in the subject's home setting. The fact that up to five subjects also sat in the same room at each visit, albeit apart from each other, could also have affected their eating behaviour.
In summary, ingestion of chocolate milk increased subjective ratings of satiety and fullness compared with cola and decreased hunger and prospective consumption, whereas ad libitum energy intake was unaltered. The present study has been conducted in the light of controversy about the role of sweetened beverages and their failure to satisfy hunger. Long-term interventions have found that consumption of sugar-sweetened sodas leads to increased energy intake and weight gain in both adults and children (Ludwig et al. Reference Ludwig, Peterson and Gortmaker2001; Raben et al. Reference Raben, Vasilaras, Møller and Astrup2002; Bray et al. Reference Bray, Nielsen and Popkin2004; Schulze et al. Reference Schulze, Manson, Ludwig, Colditz, Stampfer, Willett and Hu2004). The present study involved a small homogeneous group of subjects in a simple cross-over study. Further long-term studies are required to investigate how different beverages affect weight and appetite factors and to determine if and how differences in appetite ratings can be converted into that which is ultimately required, a difference in energy intake.
Acknowledgements
The authors thank Charlotte Kostecki and the kitchen staff at the Department for their assistance, Lone Brinkmann Sørensen for help with the statistical analysis and the subjects for their interest in the study. A. A. is a member of several advisory boards for the food industry and is a Medical Advisor for Weight Watchers. The present study was supported by Arla Foods.
