Participants
Thirty-eight healthy Caucasian participants (19 men and 19 women; age 25 ± 9 y (mean ± SD, range 18-51 y)) with a body mass index (BMI) of 25.0 ± 3.3 kg/m2 (mean ± SD, range 20.3-31.2 kg/m2) participated in this study. Based upon the study by Lemmens et al. [5], power analysis showed that with an α of 0.0125 (taking into account the Bonferroni correction for multiple testing) and β of 0.10 (power = 1-β = 0.90), at least 31 participants were needed. They were recruited by advertisements in local newspapers and on notice boards at the university. Participants underwent an initial screening including measurement of body weight, height, waist circumference and hip circumference, and completed a questionnaire related to health, use of medication, physical activity, and eating behavior. Inclusion criteria comprised BMI 20-30 kg/m2, both genders, no use of medication (except contraception), no food allergies, no dietary restrictions, and not pregnant or breast-feeding. Regarding overweight participants only participants with abdominal adiposity were included, as chronic stress has been associated with visceral fat accumulation and obesity [1, 33, 34]. Abdominal adiposity was defined as having a waist circumference of ≥ 80 cm in women and ≥ 94 cm in men [35].
Eating behavior was analyzed using a validated Dutch translation of the Three Factor Eating Questionnaire (TFEQ) which measures three components: 'cognitive restraint of eating' (factor 1, F1), 'disinhibition of restraint' (factor 2, F2), and 'hunger' (factor 3, F3) [36]. On the basis of the median for the TFEQ scores in the south of the Netherlands, participants were characterized as unrestraint when dietary restraint scores were < 9, and as restraint when scores were ≥9. Participants were characterized as having low disinhibition when disinhibition scores were < 5, and as having high disinhibition when scores were ≥5 [37].
All participants gave written informed consent and the study was approved by the Medical Ethical Committee of the Maastricht University.
Study design
The study was conducted in a randomized cross-over design. All participants came to the university four times, in a fasted state (for at least 8 h), between 08:00 and 9:00 AM. They came once for a stress session receiving a high-protein meal, once for a rest session receiving a high-protein meal, once for a stress session receiving a high-carbohydrate meal and, once for a rest session receiving a high-carbohydrate meal. The order of the four conditions was randomized across the participants to prevent any order effects. The four test sessions were at least one week apart.
Figure 1 gives a schematic overview of the study design. After arrival at the university, participants were seated in the laboratory and remained seated throughout the experiment. Each test session participants received 50 g of yoghurt ('Campina magere yoghurt naturel', 84 kJ, energy percentage protein/carbohydrate/fat (En% P/C/F) 53/44/2) to prevent large hunger feelings. The test sessions started two hours later, to give the participants the chance to adapt to the laboratory environment.
An ego-threatening computer test containing elements of an IQ-test was used to create the stress vs. rest conditions in participants [6, 38, 39]. Two versions of this psychological test were used: a difficult stress version with not enough time to solve the assignments and an easier control version with enough time to solve the assignments. The psychological test was an updated version of the test used by Rutters et al. and Born et al. [6, 30].
Following this psychological test (Figure 1, 'Rest/Stress test') food reward, i.e. 'liking' and 'wanting', was measured by means of a computer test, described and validated by Lemmens et al. [40]. During the 'liking' part of the computer test, participants had to indicate their relative preference of paired items within and between six categories: bread, filling, drinks, dessert, snacks, and stationery (non-food alternative as control). During the 'wanting' part, participants had to work to earn items by playing memory games for each of the same six categories.
After completing the 'liking' and 'wanting' computer test participants were offered the test meal (lunch), which was either a high-protein meal or a high-carbohydrate meal, and which had to be consumed entirely. Participants received an amount corresponding to 30% of their daily energy requirements.
Following test meal consumption, participants completed the psychological test and the 'liking' and 'wanting' computer test again. Subsequently, participants received and consumed their wanted food items (Figure 1, 'Wanted meal', post-meal energy intake), which were chosen by means of the 'wanting' part of the computer test.
The psychological 'Rest/Stress test' and the 'liking' and 'wanting computer test were completed pre and post test meal to be able to measure effects of stress on food choice and food intake in hunger as well as in the satiated state. During the rest session participants completed twice the control (rest) version of the psychological test, and during the stress session they completed twice the stress version of the psychological test.
To investigate whether the stress condition inflicted psychological changes, we used Profile Of Mood State (POMS) and State Trait Anxiety Inventory (STAI) questionnaires. One hundred unit visual analogue scales (VAS; in mm) were used to assess the appetite profile. Questionnaires were collected seven times per test session.
All women were tested in the follicular phase, as it has been shown that women have a higher spontaneous energy intake in the luteal phase compared with the follicular phase [11, 41].
Test meals
The test meal was either a high-protein lunch (En% P/C/F 65/5/30) or a high-carbohydrate lunch (En% P/C/F 6/64/30). Both meals were isocaloric and matched for energy density: 4 kJ/g. The amount of the meals that was given to the participants corresponded to 30% of their daily energy requirements (DER). For each participant the DER were calculated by multiplying the basal metabolic rate (BMR) by the appropriate physical activity factor (1.5-1.8, derived from the screening questionnaire, [42]). The BMR (kcal/day) was calculated according to the equation of Harris-Benedict [43].
The high-protein meal consisted of a salad (iceberg lettuce, cucumber, mushroom, and sunflower oil), Gouda cheese, salami, and a strawberry protein shake (protein source: milk protein). The high-carbohydrate meal consisted of a salad (iceberg lettuce, cucumber, green pepper, and sunflower oil), savory cheese biscuits and TUC bacon biscuits, and a strawberry carbohydrate shake (type of carbohydrate: dextrin maltose). In both meals the shakes represented 47 En% of the total meal.
During the screening session participants had to taste and rate the food items, which would have to be consumed on the test days, for subjective liking (VAS), in order to check whether all food items were acceptable. All food items scored more than 60 mm on a 100-mm VAS.
Questionnaires
One hundred unit VAS (mm) were used to assess the appetite profile. The scales were anchored with 'not at all' at one end and 'extremely' at the other end, and combined with questions on feelings of hunger, thirst, fullness, satiety, and desire to eat, and on subjective liking and wanting of the test meals.
Mood states were assessed using a modified version of the Dutch translation of the POMS [44]. This questionnaire contains 35 adjectives that are rated on a five-point scale and is divided into five subscales (depression, tension, confusion, fatigue, and anger). The Dutch translation of the state scale of the STAI questionnaire was used to measure state anxiety [45]. Participants had to rate 20 statements on how they felt at that moment on a four-point scale. An increase in POMS and STAI scores is associated with a worsening in mood.
The VAS (for hunger, thirst, fullness, satiety and desire to eat), POMS, and STAI questionnaires were completed seven times throughout the test sessions at 0, 30, 80, 115, 155, 205 and 225 minutes (Figure 1). The VAS on subjective liking and wanting of the test meals were completed pre and post test meal consumption (at 80 and 115 minutes).
'Liking' and 'wanting' computer test
The computer test described and validated by Lemmens et al. [40] was used to measure the rewarding value, i.e. 'liking' and 'wanting', for 72 items divided in six categories: bread, filling, drinks, dessert, snacks, and stationery (non-food alternative as control). Each category contained 12 items. The 72 items were presented as photographic stimuli on a computer screen (13-inch Mac Book, Apple, Cupertino, USA).
The computer test contained two parts, a 'liking' part and a 'wanting' part. Both the 'liking' and 'wanting' tasks assessed 'liking' respectively 'wanting' for the same food and stationery items. During the 'liking' part, participants had to indicate their relative preference of paired items within and between the six categories. This resulted in a relative ranking of 'liking' of the items per category and of the categories.
During the 'wanting' part, participants had to work to earn items by playing memory games. For each category of items participants played a five by five memory game (12 pairs of items) followed by the indication of the items participants wanted to acquire at that moment. The more pairs of items were found in the memory game, the more randomly selected items were offered to choose from afterwards, e.g. if eight pairs of items would be found in the memory game of the snacks category, then eight randomly selected snacks would be offered to choose from. Participants could choose zero, one or two items per category. They were instructed to choose the items while keeping in mind that all the chosen items would be offered to them and had to be eaten completely. The chosen items obtained a score equal to the number of pairs of items found in the memory game, representing the motivation or workload for the chosen items. Items not chosen obtained a score of zero. Per category the sum of the scores of the items was calculated and represented the 'wanting' score for each category. During the screening session participants were tested on their ability to play a five by five memory game within two minutes.
All the food items chosen by means of the 'wanting' part of the post test meal computer test were offered to the participants at a fixed amount, which was described to the participants beforehand, and food items were eaten completely (Figure 1, 'wanted meal'). Total energy content (post-meal energy intake) and macronutrient composition of the consumed wanted food items was calculated.
Statistics
Data were analyzed using StatView 5.0 (SAS Institute Inc., Cary, NC, USA). Unpaired Student's t-tests were used to analyze differences in participant characteristics between men and women. Factorial ANOVA with repeated measures was used to study the effects between participant groups (men vs. women, overweight vs. normal weight, high vs. low disinhibition) of the conditions of stress vs. rest and of high-protein vs. high-carbohydrate, and of time (pre and post test meal), on data of the questionnaires (VAS, POMS, STAI), on data of the 'liking' and 'wanting' computer test, and on post-meal energy intake. Paired and unpaired Student's t-tests were used as Post hoc analyses for significant interactions. Areas under the curve (AUC) for questionnaire data were calculated using the trapezoid method. All tests were two-sided and differences were considered significant at P < 0.05. Values are expressed as mean ± standard error of the mean (SEM), unless stated otherwise.
Neither gender differences (men vs. women), nor differences according to weight status (overweight vs. normal weight) were detected concerning possible conditional effects of stress vs. rest and of high-protein vs. high-carbohydrate on data of the questionnaires (POMS, STAI, VAS), on data of the 'liking' and 'wanting' computer test, and on post-meal energy intake. However, an effect of disinhibited eating behavior was detected for some of the measured variables. Therefore, results were presented firstly for all participants together and secondly for individuals characterized by high vs. low disinhibited eating behavior (high F2 score: n = 16, 6 men and 10 women; low F2 score: n = 22, 13 men and 9 women).