OBJECTIVE: To test if a diet of 4.2 MJ/24 h as six isocaloric meals would result in a lower subsequent energy intake, or greater energy output than (a) 4.2 MJ/24 h as two isocaloric meals or (b) a morning fast followed by free access to food.
DESIGN: Subjects were confined to the Metabolic Unit from 19:00 h on day 1 to 09:30 h on day 6. Each day they had a fixed diet providing 4.2 MJ with three pairs of meal patterns which were offered in random sequence. They were: six meals vs two meals without access to additional foods (6vs2), or six meals vs 2 meals with access to additional food (6+vs2+), or six meals vs four meals (6+vsAMFAST). In the AMFAST condition the first two meals of the day were omitted to reduce daily intake to 2.8 MJ and to create a morning fast, but additional food was accessible thereafter. Patients were confined in the chamber calorimeter from 19:00 h on day 2 until 09:00 h on day 4, and then from 19:00 h on day 4 to 09:00 h on day 6. The order in which each meal pattern was offered was balanced over time.
MEASUREMENTS: Energy expenditure (chamber calorimetry), spontaneous activity (video) and energy intake (where additional foods were available) during the final 24 h of each dietary component.
SUBJECTS: Ten (6vs2), eight (6+vs2+) and eight (6+vsAMFAST) women were recruited who had a BMI of greater than 25 kg/m2.
RESULTS: From experiment 6vs2 the difference between energy expenditure with six meals (10.00 MJ) and two meals (9.96 MJ) was not significant (P=0.88). Energy expenditure between 23:00 h and 08:00 h (‘night’) was, however, significantly higher (P=0.02) with two meals (9.12 MJ/24 h) compared with six meals (8.34 MJ/24 h). The pattern of spontaneous physical activity did not differ significantly between these two meal patterns (P>0.05). Total energy intake was affected by neither meal frequency in experiment 6+vs2+ (10.75 MJ with six, 11.08 MJ with two; P=0.58) nor a morning fast in experiment 6+vsAMFAST (8.55 MJ/24 h with six, 7.60 MJ with AMFAST; P=0.40). The total diet of subjects who had a morning fast tended to have a lower percentage of total energy from carbohydrate (40%) than when they had six meals per 24 h (49%) (P=0.05). Subsequent energy balance was affected by neither meal frequency (6vs2; P=0.88, 6+vs2+; P=0.50) nor a morning fast (P=0.18).
CONCLUSIONS: In the short term, meal frequency and a period of fasting have no major impact on energy intake or expenditure but energy expenditure is delayed with a lower meal frequency compared with a higher meal frequency. This might be attributed to the thermogenic effect of food continuing into the night when a later, larger meal is given. A morning fast resulted in a diet which tended to have a lower percentage of energy from carbohydrate than with no fast.
International Journal of Obesity (2001) 25, 519–528
Obesity may be inversely associated with meal frequency1,2,3,4 and directly associated with delaying energy intake over the day.5,6 Such findings are not, however, found consistently in all studies7 and because subjects were free-living, may simply reflect the degree of under-recording, as a function of degree of obesity8 or poor compliance with the study protocol.
A chamber calorimeter provides a controlled environment in which the impact of meal pattern on the two factors that contribute to obesity in the long term, namely energy intake and energy expenditure, can be measured. Others9,10,11,12 used fixed activity protocols and found that meal pattern had no impact on total energy expenditure. Dallosso et al 9 however, suggest that a fixed activity protocol may be inappropriate as it precludes the expression of spontaneous activity. They found a weight increase suggestive of more positive energy balance of 1.66 MJ/24 h when two meals per 24 h were given compared with six meals per 24 h over a 14 day period (free-living). Such a difference in energy balance was not seen when subjects followed a strict activity protocol whilst total energy expenditure was measured in a chamber calorimeter over 24 h. In this study it was thus decided that meal frequency should be manipulated whilst subjects were in a chamber calorimeter but free to select their own activity pattern.
Where a prescribed energy diet is given, the frequency of meals over which the prescribed energy is consumed might influence the selection of additional, non-prescribed foods, hence subsequent total energy intake. A chamber calorimeter provides the opportunity to manipulate the meal frequency of the prescribed foods whilst allowing subjects to self-select from an additional range of foods.
Three intervention studies were thus designed which aimed to investigate energy expenditure, spontaneous activity and additional food selection in obese subjects experiencing meal pattern manipulation in a chamber calorimeter. The meal pattern manipulations used were a comparison of six meals with two meals per 24 h (6vs2), protocol 6vs2 with access to additional foods (6+vs2+) and a comparison of a morning fast (followed by four meals with unlimited access to additional foods) with six meals per day with unlimited access to additional foods (6+vsAMFAST). The three studies were undertaken in a secure metabolic unit. Total energy expenditure was measured using a chamber calorimeter and pattern of spontaneous activity was recorded by video.
Design and methods
The studies were approved by the Ethics Committee of St Bartholomew's Hospital District Research Ethics Committee and all subjects gave written informed consent.
Subjects were recruited from an outpatient obesity clinic at St Bartholomew's Hospital (Consultant in charge, John S Garrow) and by referral from State Registered Dietitians at associated London Hospitals. Inclusion criteria were female, body mass index (BMI) greater than 25 kg/m2, over 18 y of age and able to gain potential therapeutic benefit from a measurement of energy expenditure. Exclusion criteria were use of insulin or hypoglycaemic drugs for diabetes mellitus, major dietary changes during the previous 6 months, or intolerance or dislike of more than two of the foods that would potentially be offered. Table 1 shows the mean age and BMI of the subjects in each study.
Protocol (Figure 1)
Subjects were admitted at 18:00 h to The Calorimetry Unit (a secure metabolic unit) at St Mark's Hospital, City Road, London (day 1) and were given a reheated, commercially produced, frozen meal (Menumasters chicken platter) which had to be be consumed by 19:00 h (except for chicken bones and skin). Subjects were introduced to the chamber calorimeter which would act as their room throughout their admission. Subjects were required to stay within the confines of the unit until 09:30 h on day 6, but had access to nurse call facilities at all times.
From 09:00 h on day 2 subjects were allocated randomly to one of three studies which provided 2 days of one meal pattern followed by a second 2 days of a second meal pattern. The order in which each meal pattern in a pair was offered was balanced with time.
Subjects were given 4.2 MJ as six isocaloric meals at 2 h intervals between 09:00 h and 19:00 h or 4.2 MJ as two isocaloric meals at 11:00 h and 19:00 h.
Protocol 6vs2 was repeated but subjects were given free access to a range of additional foods (Appendix 2).
Subjects were given 4.2 MJ as six isocaloric meals at 2 h intervals between 09:00 and 19:00 h with free access to additional foods or fasted until 13:00 h and then received 2.8 MJ as four isocaloric meals (13:00, 15:00, 17:00 and 19:00 h) and free access to additional foods (Appendix 2).
Subjects were confined to the chamber calorimeter for 38 h from 19:00 on the first day of each of the 2 day meal patterns (day 2 and day 4). Measurements of total energy intake, energy expenditure and spontaneous activity pattern from the final 24 h of each of these periods were used in analysis.
Ten subjects were asked to estimate their daily energy intake over the 4 day period, once all the other measurements had been made.
Provision of prescribed meals
A 2 day menu cycle was designed which consisted of six ‘meal’ units of approximately 167 kcal each (Appendix 1). The meal pattern required was constructed by grouping together or omitting ‘meal’ units. The distribution of total energy was 13% from protein, 34% from fat and 53% from carbohydrate. The menu was constructed using manufacturers' nutritional data and food composition tables.13 Where possible each food type used for one subject came from the same packet or batch throughout the feeding period. These foods had to be consumed as soon as they were given to the subject.
Provision of foods to which subjects had free access
Where the protocol required subjects to have free access to additional foods, a quantity of each product, in excess of likely consumption, was weighed and placed in the calorimeter fridge (Appendix 2). Intake was calculated by difference at the end of each period of availability using manufacturers' nutritional data and food composition tables.13 Subjects were asked to pick a diet that was as close as possible to their home diet given the constraints of the study environment. Where possible each food type used for one subject came form the same packet or batch throughout the feeding period.
Measurement of energy expenditure
The two calorimetry chambers used had volumes of 12 000 and 11 900 l, respectively. Each chamber was approximately 2.2×2.2×2 m and contained a single bed, wash basin, commode chair, desk, fridge, television, video recorder and exercise bike. Two air hatches allowed food to be passed into the chamber and excreta to be passed out. Subjects had unrestricted use of a ward nurse call facilities and could communicate via an intercom with their carers. Visiting by friends and relatives during the measurement was discouraged and verbal contact with the experimentor was restricted to 09:00, 11:00, 17:00 and 21:00 h whilst energy expenditure was measured.
Positive pressure was maintained within the chamber whilst it was ventilated with fresh air at a rate of 80 l per minute. The oxygen content of the air was measured every 6 min (Oxygen analyzer Servomex 1100A) and the data stored on a data logger (Grant squirrel data logger, Wessex Power Technology). The procedure by which energy expenditure can be measured in an open circuit indirect calorimeter from oxygen consumption only (without analysis of carbon dioxide production) is set out by Brown et al. 15 A spreadsheet for making this calculation was provided by Coyle14 (Appendix 5). By this calculation heat production (kJ/min) was 16.494 times oxygen uptake (l/min).
Calibration was undertaken using butane. Theoretically 1 g of butane uses 2.51 l oxygen for complete combustion. However the butane lamp when burned for 2 h in a Deltatrac ventilated in 13 replicate runs gave an oxygen consumption of only 2.3 l oxygen/g butane, indicating that combustion was only 91.7% complete. Using this correction factor the heat output of the butane lamp measured on six replicate runs by the calorimeter chamber A was 100.3 (s.d. 3.2)% of the theorectical value and in chamber B was 99.1 (s.d. 5.3)% of the theoretical value.14
Intra-subject variation had been obtained by Coyle.14 Twenty-one subjects spent two consecutive days in the calorimeter d followed a strict activity protocol. Mean total energy expenditure was 7.8 MJ/24 h day 1 (s.d. 1.4) and 7.7 MJ/24 h (s.d. 1.5) on day 2. The mean difference was −0.09 MJ/24 h (s.e. 0.13 MJ) which was not significant (P=0.21). The 95% confidence interval for the difference over repeated measures was −0.36–0.18 MJ.
Measurement of spontaneous activity pattern
Activity patterns were recorded by a video recorder (1.25 pictures per second). Mirrors allowed activity in all parts of the chamber to be seen.
On replaying the videos four distinct categories were identified. These were cycling, standing/walking, sitting on the chair/sitting on the bed and flat on the bed seemingly asleep. The frequency and time period of each of these activities was logged. Subjects were asked to give written informed consent to this aspect of the study as a separate component. They were informed that refusal would not mean their exclusion from the remainder of the study.
Statistical analysis and calculation of power
Paired data was compared using the Wilcoxon matched-pairs signed-ranks test (two-tailed, SPSS PC statistics package, version 5.0.1). It was estimated that 10 subjects were sufficient to detect a difference of 0.84 MJ in energy expenditure with more than 80% (PowerPack' version 2.2) when prescribed foods alone were consumed (6vs2). The estimation was based on the intra-subject variation previously noted in the calorimeter14 and work by Dalosso et al.9 The latter suggested that subjects might have a positive energy balance of 1.66 MJ/24 h when they were given two meals per day compared with six meals per day. Power calculations were not undertaken for other aspects or when additional foods were given because there was no previous data available on which they could be based.
Ten 6vs2, eight 6+vs2+ and eight 6+vsAMFAST subjects were recruited. Weight, height and BMI are shown in Table 1.
Neither meal frequency nor a period of fasting resulted in significant differences in total energy expenditure (Table 2: 6vs2, P=0.88; 6+vs2+, P=0.61; 6+vsAMFAST; P=0.13). ‘Night’ energy expenditure was significantly higher (P=0.02) with two meals per day compared with six meals per day when prescribed foods alone were given (6vs2, n=8).
Standing frequency alone (but not duration of activity P=0.26) was increased significantly with six meals per day compared with two meals per day (6vs2, P=0.02; Table 3). This was attributed to subjects standing on four extra occasions to collect meals from the hatch. Only four subjects consented to measurement of spontaneous activity in 6+vs2+ and 6+vsAMFAST so no statistical analysis was undertaken. No major intra-subject differences were noted with different meal patterns.
Total energy and macronutrient intake
Total energy intake (Table 2) was higher (not significant, P=0.58) with two meals per day compared with six meals per day (6+vs2+). Fasting in the morning (6+vsAMFAST) in fact resulted in a non-significant reduction (P=0.40) in total energy intake compared with six meals per day with additional access to foods. In all cases a range of foods was selected (Table 4). At an individual level, the greatest additional intake of energy from one food was 4.7 MJ from cheddar cheese (in a subject who had been fasting). The morning fast tended to result in the selection of additional foods that resulted in a lower percentage of energy from carbohydrate in the final diet compared with the final diet with six meals per day (P=0.05; 6+vsAMFAST; Table 2). The converse was seen for fat and protein but these were not significant differences (P=0.12, P=0.40).
Neither meal frequency (6vs2, P=0.88; 6+vs2+; P=0.50) nor a morning fast (P=0.18) resulted in differences in energy balance that were statistically significant (Figures 2, 3 and 4). All subjects were in negative energy balance when they were given 4.2 MJ as six or two meals per day (6vs2; Figure 2). Subjects tended to achieve positive energy balance when they were given free access to other foods in addition to 4.2 MJ as six or two meals per day (6+vs2+; Figure 3). A morning fast resulted in a range of outcomes relative to six meals per day with access to additional foods (6+vsAMFAST; Figure 4). Four subjects were in more negative energy balance with a morning fast with three achieving actual negative energy balance. Three subjects were in more negative energy balance with six meals per day than with the morning fast and two of these achieved actual negative energy balance.
Subject estimation of energy intake
Table 5 shows that the mean difference between the subject estimate of energy intake and the measured energy intake was an underestimate of 1.3 MJ/24 h. One subject correctly estimated their intake, four subjects made on overestimate and five subjects underestimated their actual intake.
These three studies aimed to investigate the impact of meal frequency manipulation and a morning fasting on total energy expenditure, spontaneous patterns of activity and ad libitum food selection in obese subjects in a chamber calorimeter. Energy expenditure was measured by 24 h chamber calorimetry, and patterns of spontaneous activity were measured by video recorder.
When 4.2 MJ/24 h was given as two meals per day or six meals per day there was found to be no significant difference in total energy expenditure. A range of spontaneous activity patterns were seen despite the constraint of the chamber calorimeter but although there was variation between subjects there was no significant within-subject difference with meal frequency. An apparent increase in standing with six meals per day could be attributed to getting up to collect the extra four meals from the hatch compared with two meals per day. We have thus failed to support the suggestion made by Dallosso et al 9 that differences in spontaneous activity might account for more positive energy balance with 2 meals per day compared with 6 meals per day. Work by Verboeket van de Venne and Westerterp11 suggest that this may also be true in the longer term in energy-restricted subjects.
The video recordings had insufficient clarity to detect whether non-exercise activity thermogenesis (NEAT) such as fidgeting were changed. Ravussin et al 16 suggest that such activity could account for as much as 0.42–3.36 MJ/24 h. Levine et al 17 suggest that between-subject differences in this component of energy expenditure might be an important determinant of fat deposition when a hyperenergetic diet is given. In this case, the failure to find a significant difference in total energy expenditure with meal frequency suggests that any change in NEAT was small or there was a concurrent change in diet-induced thermogenesis or basal metabolic rate which is not supported by the literature (diet-induced thermogenesis;11,18 resting metabolic rate19).
We did, however, find a lag in energy expenditure with two meals per day compared with six meals per day demonstrated by the significantly higher energy expenditure during the ‘night phase’. This was in agreement with Dallosso et al.9 This might be attributed to the delayed thermogenic effect of the three meal units given at 19:00 h compared with the more evenly spaced meal pattern with six meals per day. This had no impact on total energy expenditure so we must question whether there might be a consequence to the other side of the energy balance equation—energy intake. The internal milieu during this period of post prandial thermogenesis would be of the fed state. Heini et al 20 demonstrate that satiety was positively associated with the blood glucose (P=0.02) and insulin response (P=0.020) to a test meal. In the fed state the individual is exposed to fewer internal food cues so must only contend with external food cues. When this ‘protection’ occurs at night it is potentially wasted. On the other hand, with six meals per day, this post-prandial phase occurs to a greater extent whilst the subject is awake and has more ready exposure and access to food. Verboeket van de Venne and Westerterp10 also note that two meals per day result in a strong diurnal periodicity in fat and carbohydrate oxidation compared with a nibbling pattern. They conclude that this has no impact on total energy expenditure but it may perhaps impact upon physiological cues for energy intake. Bellisle21 concurs with this by concluding that any impact of meal frequency on energy regulation are likely to be mediated via the energy intake side of the energy balance equation.
Study 6+vs2+ allows us to test this hypothesis as subjects again were given 4.2 MJ/24 h as either six or two meals per day but in addition they had ad libitum access to a range of additional foods. Six meals per day did result in a lower total energy intake than with two meals per day but the difference is not significant. We also find that there is no difference in patterns of energy expenditure during the ‘night’ period. This might either be because this was a phenomena unique to those subjects who participated in the previous study and who were in fact a younger group. Alternatively subjects may have selected foods in such a way that they smoothed out the differences induced by the prescribed food. The video quality did not allow time of eating and selection of food to be logged so only total daily consumption can be considered.
The final study (6+vsAMFAST) aimed to mimic a patient missing breakfast and fasting until lunchtime. We found no evidence of subsequent over-compensation, indeed the marginal increase in additional foods consumed after the fast was insufficient to compensate for the fast so the total intake for the day was lower (not significant). After the fasting period subjects selected additional foods which resulted in a dietary composition which contained a lower percentage of carbohydrate than the diet selected with six prescribed meals per day (P=0.05). The latter had been similar to when six meals were given per day in 6+vs2+ an identical intervention but with different subjects. This suggests reproducibility when selecting composition although less reproducibility was noted in total energy intake relative to expenditure (Table 2 and Figures 2 and 3). When subjects in 6+vsAMFAST were given 4.2 MJ/24 h as six meals they selected less additional energy than the group of subjects experiencing the identical protocol in 6+vs2+. This is perhaps a function of being a younger, more obese group or maybe due to a characteristic not considered such as restraint score.
The difference in composition of the diet selected might have an impact on the longer term energy intake regulation by influencing the control of the size of an eating episode (satiation) and the strength of post-ingestive appetite inhibition (satiety). A diet with a higher proportion of carbohydrate might be expected to induce greater satiation and satiety than one with a lower carbohydrate content,22 especially if it was less energy dense,23 foods selected were starch based24 and in comparison with a diet in which a higher proportion of energy was from fat.25 Over the longer term this might bestow an advantage on a patient attempting to reduce their energy intake.
Selecting a diet lower in carbohydrate after a fast compared with no fast is contrary to the expectation that they might have experienced a greater depletion of glycogen stores, hence select carbohydrate-rich foods. The prescribed diet had a relatively high percentage of energy from carbohydrate compared with the UK population. The lower percentage of carbohydrate with a higher proportion of self-selected foods may reflect the greater influence of habitual intake patterns.
An additional, striking finding of these studies is that, irrespective of meal frequency, subjects are poor at selecting a diet which complies with their dietary prescription. All reported following a weight-reducing diet at home, yet despite being asked to eat as at home, in many instances, achieved positive energy balance whilst in the chamber calorimeter. This might either be attributed to the unnatural situation of a calorimeter or ignorance of appropriate food selection. Interestingly the foods selected were often of a type which might be considered ‘good’ but were consumed in quantities which resulted in positive energy balance. Some subjects were certainly poor at estimating their intake when asked (Table 5). This has important implications for diet therapy as it indicates patients attempting to lose weight need information about appropriate quantities as well as food type.
These three studies do not support the hypothesis that a lower meal frequency, when compared with a higher meal frequency might result in more positive energy balance by inducing a lower energy expenditure or a higher energy intake in the short term. A period of fasting does not result in over-compensation, but in fact results in a lower total energy intake (non-significant). A morning fast resulted in a lower percentage of total energy from carbohydrate than with a nibbling pattern, which might have implications for energy intake in the longer term. Future work is required to explore the reproducibility and importance of this finding.
We are grateful to the staff of Salmon and Holgate Wards, St Mark's Hospital for patient care and to the patients for their participation. We are grateful for statistical advice from The Biometrics Department at St Bartholomew's Hospital Medical College (Peter Brown, Janice Thomas and Kay Sanders), Malcolm Law and John Stapleford.