Obtaining accurate estimates of energy intake (EI) in the free-living population is a difficult proposition(Reference Macdiarmid and Blundell1, Reference Lara, Scott and Lean2) as there are extensive limitations associated with all available methods. As a result, direct, laboratory-based observation of EI has become increasingly common, particularly in short-term exercise intervention studies which include buffet meals to assess EI(Reference George and Morganstein3–Reference Martins, Morgan and Bloom5). Conclusions in these studies are thus often drawn from a single measurement obtained in an atypical setting. Reproducibility of EI measurements made using this method has been shown to be high for sedentary and moderately active lean males at rest(Reference Arvaniti, Richard and Tremblay6–Reference Nair, Brennan and Little8); reproducibility coefficients range from 0·86 to 0·97. The only existing study assessing reproducibility of post-exercise EI reported a coefficient of 0·89 obtained from male and female recreational athletes(Reference Laan, Leidy and Lim9). All of these existing studies assess EI at a single meal only; none has examined the reproducibility of 24-h EI in overweight, sedentary females either under sedentary or under exercise conditions. Thus, it is not known whether findings in these short-term exercise studies are exercise-induced or simply attributable to day-to-day variation in food intake. Women, particularly overweight and obese, are known to frequently underreport their dietary intake(Reference Macdiarmid and Blundell1); hence EI assessment is perhaps most difficult in this group. Reproducibility of this method in lean males cannot be assumed to be similar for overweight and obese women due to differences in adiposity, and greater hormonal influences in women(Reference Barr, Janelle and Prior10). The present study aimed to examine the reproducibility of acute post-exercise EI responses, measured via ad libitum buffet style meals over a 24-h period, in overweight, pre-menopausal, sedentary females.
Subjects and methods
Subjects
This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human subjects were approved by the University of Glasgow Medical Faculty ethics committee. Written informed consent was obtained from all subjects. Subjects were recruited by poster and e-mail advertisements around the University. A total of fourteen healthy women gave written, informed consent to participate and all completed the study. Subjects were pre-menopausal, healthy, non-smokers, currently sedentary (participation in < 2 h planned exercise/week), not currently attempting weight loss, and were classed as either overweight or obese (BMI ≥ 25 kg/m2).
Screening
All subjects completed standard questionnaires to assess their current physical activity levels (International Physical Activity Questionnaire), health status and food preferences before beginning the study. Height was measured to the nearest cm using a stadiometer (Seca 213; Seca, Birmingham, UK). Body mass (to the nearest 0·1 kg) and composition were measured using foot-to-foot bio-impedance scales (Tanita B.V., Hoofddorp, The Netherlands).
Sub-maximal fitness test
Before the beginning of the trial, subjects completed a graded, sub-maximal treadmill fitness test. Tests were conducted according to exercise testing guidelines(11) and began at a moderate walking speed which was tailored to an individual subject's fitness level, and the speed was gradually increased until the heart rate reached 85 % of the age-predicted maximum. Each stage lasted 5 min and the heart rate and expired air samples were taken in the last minute of each stage to enable the prediction of maximal oxygen (VO2max) consumption.
Study design and measures
Each subject completed two sets of trials; hence participation involved a total of four trials – two exercise and two control. Subjects were asked to avoid alcohol and standardise their food intake before each trial. All trials were timed to control for effects of the menstrual cycle for each subject; the trials were carried out at least 3 weeks apart in practice. The first set of trials, one exercise and one control, was completed in a randomised, counter-balanced fashion. The second set of trials was completed in reverse order so as to minimise potential bias effects of order.
Each trial lasted 24 h, spanning over 2 d; observation was carried out in the afternoon of day 1, and in the morning of day 2. Subjects attended the laboratory on day 1 at 14.00 hours and remained for 4 h, during which time the intervention (exercise or control) period was completed. Subjects fasted overnight at home and returned to the laboratory the next morning at 09.00 hours, remaining for a further 5 h. Body composition was measured via bio-impedance in the fasted state at the beginning of day 2 of all trials. During each trial, three ad libitum buffet meals were served; evening meal on day 1, and breakfast and lunch on day 2.
Intervention
During the exercise trials, a moderate (65 % VO2max) intensity treadmill walking session was carried out in the afternoon of day 1 to expend 1·65 MJ; an energy expenditure similar to recommendations for individual sessions of exercise aimed at weight control(Reference Donnelly, Blair and Jakicic12). At all other times, the subjects were sedentary. The control trials were identical, except that the subjects remained sedentary during the intervention period.
Ad libitum buffet meals
Buffet meals included a selection of commercially prepared foods which were identical between trials; all meals included a selection of fresh, chopped fruit, orange juice and yoghurt, while lunch and evening meals both included salad with dressing served separately and a selection of cupcakes. Foods served specifically at certain meals included cereal, milk, croissants and jam at breakfast, soup and a selection of sandwiches at lunch, and a pasta ready meal at the evening meal. Exact individual menus were tailored to each subject's personal tastes to ensure that all foods presented were palatable. The amount of food presented at each meal was significantly more than the subject was expected to consume to ensure that it was possible to eat to satiety without requiring additional food to be presented mid-meal. Subjects were left alone for 30 min to consume each meal. All foods were weighed before and after consumption to enable the calculation of EI using nutrient analysis software (Windiets 2005; Robert Gordon University, Aberdeen, UK). Subjects were not informed that EI was being monitored, in an attempt to minimise potential adverse effects on consumption(Reference Herman and Polivy13).
Statistical analyses
Before analysis, all data were found to be normally distributed using the Anderson–Darling normality test. Paired t tests were used to compare characteristics of the exercise sessions. Repeated-measures ANOVA with post hoc Tukey test was used to assess differences in energy and macronutrients between trials.
Reproducibility of appetite scores and EI measurements made in identical trials were assessed using two-way mixed-effects intraclass correlation coefficients (r i). Statistical significance was set at 5 %. Data were analysed using Minitab 14 (Minitab Limited, Coventry, UK) and IBM SPSS Statistics 19 (IBM Corporation, Somers, NY, USA).
Results
Subjects' characteristics
A total of fourteen subjects completed the study with no attrition. Subjects had mean age of 35·7 (sem 2·3) years, height 162 (sem 1·5) cm, body mass 78·6 (sem 3·8) kg, BMI 30·0 (sem 1·4) kg/m2, body fat 38·7 (1·5) % and VO2max 30·9 (sem 1·9) ml/kg per min. There were no significant differences in body composition between trials.
Exercise responses
Mean energy expenditure of the exercise sessions was 1·65 (sem 0·01) MJ. There were no significant differences in intensity (Δ 3·9 (sem 4·2) % VO2max), duration (Δ 0·1 (sem 1·9) min or energy expenditure (Δ 0·01 (sem 0·05) MJ) of the two exercise sessions conducted during the exercise trials.
Energy and macronutrient intake
Repeated-measures ANOVA comparison of EI in all trials showed no significant main effect of trial (P = 0·21), or interaction effect of meal × trial (P = 0·73); however, there was a significant main effect of meal; EI was significantly lower at the breakfast meal compared to the evening meal in all trials (P = 0·0003).
There was no significant main effect of trial or interaction effect of meal × trial on carbohydrate, fat or protein intake (all P < 0·01), though a significant effect of meal was observed for all macronutrients (all P < 0·0001). EI and macronutrient intake at all buffet meals are summarised in Table 1.
(Mean values with their standard errors)
a,b,c Mean values with unlike superscript letters were significantly different (effect of meal; P < 0·05; repeated-measures ANOVA).
The r i values representing the reproducibility of exercise-induced changes in energy and macronutrient intake measured at each of the three buffet meals are presented in Table 2.
(Mean values and 95 % confidence intervals)
CHO, carbohydrates.
Discussion
A substantial field of literature has built up using a potentially flawed study protocol to investigate factors which may mediate in the regulation of appetite post-exercise; the reliability of this method has not been demonstrated. The present study was designed to evaluate the reproducibility of EI values obtained from laboratory-based buffet meals, in overweight and obese women under control and exercise conditions. No acute exercise-induced compensatory increase in EI was observed in these women; however, despite the apparent agreement in measurements, the associated r i values indicate low reproducibility of these values and a large degree of individual variation associated with the EI estimates obtained via this method. The r i for exercise trial EI was not significant, which indicates very low reproducibility of this method. This finding disagrees with a previous study assessing post-exercise EI(Reference Laan, Leidy and Lim9); this study observed EI following a short exercise session (35 min) in male and female lean recreational athletes at a single non-buffet type meal, hence subjects and methodology differ significantly from the present work. r i for the difference in EI between sets of exercise and control trials was calculated to assess the reproducibility of the direction of change in EI; this was non-significant, indicating that this change was not uniform or consistent. These differences may reflect normal daily variation in EI(Reference McKiernan, Hollis and Mattes14), or indicate that the effects of exercise on EI are highly inconsistent.
The r i for EI in control trials was significant but low; the value of 0·50 represents a low level of agreement and rather weak reproducibility. The CI associated with this value must also be taken into account(Reference Hebert and Miller15). Spanning from 0·03 to 0·80, this CI reflects great individual variability in the extent of agreement and thus EI assessed in control conditions by this method will not be reliable for all subjects. This finding disagrees with similar studies involving lean males(Reference Arvaniti, Richard and Tremblay6–Reference Nair, Brennan and Little8); all of which report high reproducibility of EI values obtained from lean males at a single meal, often consisting of a single foodstuff. These studies do not report CI of reproducibility coefficients and thus the extent of variability associated with these estimates is unclear. No other study assessing the reproducibility of 24-h EI assessed via buffet meals could be identified; the EI of lean males at multiple meals could prove to be more variable than the reported values suggest. Differences in study populations, mainly in sex, adiposity and activity levels of subjects, may also contribute to the discrepancies in results between previous literature and the present study.
To our knowledge, there is no evidence regarding the reproducibility of this method in overweight and obese females with which to compare these results. This study is novel in examining this reproducibility of 24-h EI measures using this laboratory-based buffet meal method. It does not give reproducible and reliable results in overweight and obese women. It may be that short-term exercise studies may be attempting to measure a phenomenon that simply does not exist in these subjects. Further investigation is needed to clarify if this method is reliable for use in other population groups.
Like other available methods, the laboratory-based buffet meal method also has inherent limitations which may also have contributed to poor reproducibility. Food is served in an unnatural environment and this may affect eating behaviour(Reference Herman and Polivy13). Although subjects are not informed that EI is being measured, the presence of a researcher serving food may introduce perceived social pressure which can affect EI(Reference Herman and Polivy13, Reference de Castro, Brewer and Elmore16). Additionally, restrained eating status was not measured in this group of subjects, which may serve as a limitation in the present study.
Due to the exploratory nature of this study, a prospective power calculation was not carried out. However, we consider fourteen subjects to be a relevant sample size, as this number is typical of many short-term exercise studies(Reference George and Morganstein3, Reference Martins, Morgan and Bloom5) and previous reliability studies with lean males(Reference Arvaniti, Richard and Tremblay6, Reference Nair, Brennan and Little8).
This study has shown that the laboratory-based ad libitum buffet meal method does not provide a reliable, reproducible estimate of EI in all pre-menopausal, overweight and obese women under control or acute exercise conditions. The reproducibility of this method varies greatly between individuals under control and in particular post-exercise conditions; as a result, this method does not seem appropriate to assess the effects of exercise on EI in overweight women.
Acknowledgements
The present work was supported by a University of Glasgow Faculty of Medicine Studentship. The authors thank the volunteers for taking part in this study and Dr David Young for statistical advice. G. L. B. was involved in the study design and coordination, data and statistical analysis, and drafting of the manuscript; C. R. H. and M. E. L. contributed to the study design and coordination and drafting of the manuscript. None of the authors has a conflict of interest to declare.
