OBJECTIVE: The relationship between eating frequency and body fatness was tested in a population sample.
DESIGN: A cross-sectional survey on cardiovascular risk factors and a nutritional survey were carried out from June 1996 to April 1997.
SUBJECTS: Population sample of 330 free-living middle-aged men (45–64 y).
MEASUREMENTS: Body mass index (BMI), waist-to-hip ratio (WHR) and nutritional survey (3-day record).
RESULTS: In the whole sample, BMI and WHR decreased significantly (P<0.05) along with the increase of the number of eating occasions. When low energy records were excluded, the trend for BMI and WHR according to eating categories remained significant. For WHR, averages were 0.98, 0.95, 0.94 and 0.93 for 1–2, 3, 4 or 5 or more feedings a day, respectively. For BMI, mean values were 28.1, 26.2, 26.2 and 24.5 kg/m2, respectively. After adjustment for confounders (total energy intake or macronutrients, age, educational level, smoking habits, physical activity and restrained diet), the linear trend for BMI and WHR throughout feeding categories was significant when the whole sample was considered. This relationship remained similar when low energy records or when dieters were excluded.
CONCLUSION: These results suggest that for an isoenergetic intake the increase of eating frequency is associated with lower body fatness.
In the two past decades, the recent and dramatic increase in prevalence of obesity in both developed and developing countries implies that the origin of this public health problem is complex.1,2,3 The long-term persistence of energy imbalance between energy intake and reduction of physical activity, the increasing of energy density of the food and the macronutrient composition of the diet have induced weight gain, but environmental and behaviour changes, especially nutritional behaviours, may also be accountable for the increase of overweight and obesity in populations. At the end of the eighties, the average prevalence of obesity (body mass index ≥30 kg/m2) in middle aged (35–64 y) European populations was about 15% in men and 22% in women.4 The results of epidemiological and clinical investigations focusing on the influence of eating pattern and more specifically of eating frequency on body weight have been discussed.5,6 Most of interventional studies failed to show any influence of eating frequency on weight loss in dieting subjects. The studies of total energy expenditure failed to show any influence of eating frequency on energy expenditure and the reports of the thermic effect of feeding were not conclusive. Only a few epidemiological studies reported an inverse relationship between eating frequency and body weight.7,8,9,10 Methodological drawbacks have been put forward to explain some differences observed between studies or the lack of such a relationship: under-reported food intake, especially among the obese, various definitions of eating occasions, and the fact that similar studies did not take into account factors related to energy expenditure. The goal of this study was to investigate the relationship between the number of eating occasions and measures of body fatness.
Material and methods
Population and sampling
A cross-sectional survey on cardiovascular risk factors and nutrition was carried out from June 1996 to April 1997. A sample of 330 middle-aged men (45–64 y) living in the region of Toulouse (south-west of France) was recruited. The sample was selected at random from the polling lists (an exhaustive and nominal list for French inhabitants aged over 18 y) available in each town hall. The response rate reached 60% of the people contacted. Subjects were screened in a health center administered by the social security system. Participants were volunteers and received no compensation. Subjects were informed of the aim of the study and a formal consent form was completed and signed by each subject. Authorization from the appropriate ethics committee was obtained. Subjects were contacted by letter. Those who agreed to participate were given a morning appointment and asked to fast for a period of 10 h minimum.
Subjects were screened in the morning for cardiovascular risk factors by trained research nurses. Anthropometric measurements including weight, height, waist and hip circumferences were measured with standardized procedures. Body mass index (BMI) and waist-to-hip ratio (WHR) were calculated as follows: weight (kg)/height (m2) and waist/hip.
Questionnaires and personal previous history
Questionnaires were administered by interviewers at the examination center. Information on demographic and socio-economic factors, educational level and years of schooling, occupational activity, personal previous medical history, smoking habits (including past and current behaviour and the number of cigarettes, pipes or cigars smoked), drug intake and medical problems were collected. Hypocholesterolemic, antidiabetic, hypotensive or slimming diets were recorded.
Physical activity was assessed
Three levels of leisure time physical activity were defined as follows: low level (no physical activity at all), medium level (intense physical activity for 20 min once or twice a week) and high level (intense physical activity for 20 min at least three times a week). Physical activity during working hours (sedentary or standing occupation, handling objects of various sizes and weights, <10, 10–25, ≥25 kg) and commuting (duration of walking, cycling) were registered.
Food intake was assessed using a food record method.11 Since the cardiovascular risk-factor screening was completed, men received information about dietary survey and were asked if they agreed to participate. Participants were given oral and written instructions by the nurse on how to keep a three-consecutive-day food intake. Participants recorded all the food and beverage intakes (type and amount) throughout the three following days in a diary. The time, the duration and the place of food and beverage consumption were recorded too.
Three-day food records were completed using estimated weights of food, household measures and portion size. A few days (from 2 to 4 days) after the record completion, the subject was interviewed at home by a certified dietician, in the presence of the person who prepared the meals. Contents of household measures were evaluated by the dietician with a measuring glass. Food estimates were facilitated by the use of photographs showing portion sizes and their respective weights. Recorded data were carefully checked by the dietitian who, in order to avoid forgotten or misreported data, submitted a list of various food categories to the subjects to check the reliability of the data. For meals that were not taken at home (in restaurant, canteens, etc.), the cook was contacted, and the composition of the receipe and the portion size were recorded. Food data were translated into nutrient values using Renaud and Regal food composition tables.12,13
Definition of eating pattern
The number of eating occasions was counted every day and the eating frequency was calculated using the average of the three days. The value of the mean was rounded to the nearest of integer number as follows: if the mean was 4.3 the frequency was 4 and if the mean was 4.6 the frequency was 5. The first eating occasion was defined as food intake occurring after getting up (no shift worker was included in the sample). To be considered as distinct, two consecutive food intakes had to occur either in a different place or after an interval of 1 h if taken in the same place. The time interval was defined as the space of time elapsed between the last mouthful of the last food intake and the first mouthful of the following one. Any food or drink intake providing energy was taken into account. Because of the small number of subjects eating less than twice a day or more than 5 times a day, four groups of eating frequency were considered. The first one included subjects feeding 1–2 times a day, in the second and the third group the frequency of eating was 3 and 4, respectively, and in the fourth group the number of eating occasions was 5 or more.
The validity of energy intake was assessed by using the ratio between recorded energy intake and estimated basal metabolic rate.14 The basal metabolic rate was estimated according to the Schofield equations based upon body weight, age and sex.15 Multivariate analyses were performed first on the whole sample, second, after the exclusion of subjects with a ratio <1.05 and finally after the exclusion of subjects following a restrained diet.
Statistical analyses were performed using the SAS statistical software release 8.0.16 The statistical significance of the differences of BMI and WHR between the classes of environmental variables was tested by the one-way ANOVA. Pearson correlations or Spearman rank correlations when necessary were performed to test the relationship between the measures of body fatness and nutritional variables. For variables with skewed distribution, analysis was performed after logarithmic transformation. The contribution of eating frequency on the measures of body fatness was tested with a multivariate linear model. The analysis was performed after adjustment for confounding variables.
The mean value per day of the number of eating occasions was: 3.7 (mode, 3; median, 3; maximum, 8; and minimum value, 1). The majority of meals (84%) were eaten at home. (other dietary intakes were distributed as follows: in restaurants 4%, at work 4%, canteens, cafeterias 3% and various other places 5%). The average duration of the first food intake after getting-up was 12±6 min and was eaten around 7 am. The second highest energy intake was eaten around 1 pm and the average duration was 35±13 min. The third highest energy intake was around 8 pm and the average duration was 37±14 min. Table 1 gives the characteristics of the study population.
Table 2 shows the trend of the mean values of WHR and BMI according to the number of eating occasions. In the whole sample, WHR and BMI decreased significantly along with the increase of the number of eating occasions. When low energy records or when men who met dietary goals were excluded the trend remained significant for the two measures of body fatness. The reduction of the number of records was globally 10% when low energy records were excluded and 24% when dieters were excluded. The reduction of the number of records was proportionally similar in each group. When low energy records were removed it was 13, 10, 10 and 7% (χ2=0.43, NS) from the group with 1–2 feedings to the group with five feedings or more, respectively. When dieter records were removed it was 27, 25, 23 and 26% (χ2=0.17, NS) from the group 1–2 feedings to the group with 5 feedings or more, respectively.
Table 3 presents the associations of WHR and BMI with age, educational level, physical exercise, smoking habits and restrained diet. A positive and significant association was observed between WHR and age but not for BMI. A negative and significant relationship was observed between measures of body fatness and educational level or physical exercise. Men who had never smoked had a lower WHR or BMI than current smokers or former smokers. A restrained diet was associated with the highest WHR and the highest BMI.
Table 4 shows the averages of energy and macronutrient intakes according to the frequency of eating occasions. The mean values of energy were almost identical whatever the feeding categories. The exclusion from the whole sample of low energy records or dieters did not change the signifiance of the statistical test (data not shown). There was a significant trend by feeding category when macronutrients were expressed as a percentage of energy intake, and these trends were positive for carbohydrate and negative for both protein and fat.
Tables 5 and 6 give regression coefficients of the multiple regression analysis for BMI and WHR. Three models of analysis were performed. In the whole sample, after adjustment for age, physical exercise, education level, smoking habits, dieting and total energy, BMI and WHR were negatively and signicantly associated with eating frequency. After excluding from the whole sample low energy records or dieter records, associations between BMI or WHR and eating frequency were not substantially different. When both low energy and dieter records were excluded similar results were observed (data not shown). Physical exercise and educational level remained negatively and significantly associated with BMI. Age remained positively and significantly associated with WHR, whereas physical exercise was negatively and significantly associated with WHR. Former smokers had a significantly higher BMI and WHR than never smokers. The coefficient of multiple determination for WHR improved slightly when dieters were excluded. In the whole sample, the coefficient of multiple determination was 0.18, and reached 0.22 when dieters were excluded.
Multiple regression analyses for WHR and BMI were performed when total energy intake was replaced by macronutrients in the models (adjustment for the same confounders). Only carbohydrates (100 g/day) were negatively and significantly associated with BMI (β, −1.207, s.e. 0.329 and P<0.001) and the relationship between eating frequency and overweight indexes remained significant. For BMI, when saccharides were replaced by both oligosaccharides (100 g/day) and polysaccharides (100 g/day), β estimators were −1.214 (0.393) and −1.122 (0.632), respectively, and for polysaccharides P<0.01.
The overall prevalence of obesity, using the WHO recommended 30 kg/m2 cut-off point, in this population sample of men aged 45–64 was 14%. This percentage was similar to the percentage observed in other European populations.4,17 In this free-living population of middle-aged men the increase of eating frequency was associated with lower BMI and lower WHR. Adjustment for confounders (physical activity, smoking habits, age, educational level, restrained diet and energy intake) did not alter these relationships. Additional statistical analyses performed when low energy records or when dieter records were excluded from the whole sample did not change the significant and negative trend of BMI or WHR throughout eating frequency categories.
A possible bias due to misreported (under-reported) energy intake, especially among overweight men and dieters, may have altered the relationships between eating frequency and overweight indexes or even led to spurious relationships.5 This potential error was evaluated through the comparison of mean values of energy recorded throughout all the categories of eating frequency in the whole sample, after excluding low energy and dieters' records.14,18 In the whole sample, mean values of total energy were similar in all the categories of eating frequency as shown previously. When low energy records were excluded, mean values of energy ranged from 10.42 MJ/day in the three-meals group up to 10.54 MJ/day in the five-meals or more group. When dieters were excluded, mean values of energy ranged from 10.10 MJ/day in the three-meals group up to 10.40 MJ/day in the five or more meals group. When both dieters' and low energy records were excluded, energy ranged from 10.40 MJ/day in the five or more meals group up to 11.02 MJ/day in the one to two meals group (data not shown). Likewise, the proportion of low energy records was higher in men eating once or twice a day than the other frequency groups, the differences of energy between groups remaining minimal and non-significant whatever the category of records excluded. When the same selection was applied for the multivariate models the relationship between BMI or WHR and eating frequency was not modified significantly.
Furthermore, adjustment for macronutrients (fat, carbohydrates and protein) expressed as absolute value or as a proportion of energy intake did not suppress the inverse relationship observed between eating frequency and BMI or WHR. The findings of the present study tend to support the hypothesis that, for an iso-energetic intake, the number of meals per day is negatively associated with body fatness.
Some limitations of our study should be taken into consideration. A selection bias may have impaired the relationship between eating occasions and obesity when the participation rate of obese subjects was too small when compared with non-obese subjects or when subjects had modified their eating habits. The first assumption seems irrelevant since the prevalence of obese subjects in our population sample was similar to the prevalence observed in comparable European populations. In the second assumption, this bias cannot be entirely rejected, even if the observed relationship between eating occasions, BMI and WHR remained after the exclusion of subjects who had restricted their food intake. The lack of significant relationship between physical activities at work or to go to work and body fatness could be due to a non-quantitative evaluation of energy expenditure. Very moderate physical activities at work, the small proportion of subjects whose work requires intense physical activity and probably physically adapted occupations (healthy worker effect) may account for non-significant relationships.
It has been suggested that low eating frequency may be a response rather than the cause of obesity due to active restriction of food intake. Subjects may modify the pattern of eating frequency in response to their body weight changes.19 The cross-sectional design of the study did not allow any firm conclusion to be drawn about the causal relationship despite the negative relationship between eating frequency and body fatness, which remained significant when low energy and/or dieters' records were excluded from the analysis.
Similar results have already been reported.9,10 However some study have failed to show any effect of eating frequency on body fatness20 and for one prospective study the results of the cross-sectional analysis and the follow-up analysis were contradictory.8 The differences observed between the various studies might be due to the definition of meal frequency. Actually, the numerous methods used to assess the differences between a meal and a snack could lead to a different classification of the subjects and may influence the outcome of the studies. Moreover, a potential misreporting bias concerns the assessment of eating frequency habits6 and the assessment of food intake. We are aware of the fact that the reliability of food estimates increases along with the increase of the number of days of survey recorded.21 In our study we have counted the number of meals taken from one getting up to the next one and no cultural or social definition of eating occasion had been taken into account and no nutritional or food composition defining the dietary intake was used. Nevertheless, the time interval to differentiate one eating occasion from the other was arbitrary. We were aware that the outcomes were certainly time interval dependent. Another definition of the time lag between two eating occasions may have led to an increase in the number of meals if the time interval had been shorter, and to a reduction of the number of meals for subjects with the highest frequencies.
On the one hand, an inverse relationship between carbohydrates and BMI, even after adjustment for confounders, was observed for carbohydrates expressed either in grams per day or as a percentage of energy. The relationship was identical when oligosaccharides and polysaccharides were taken into account instead of carbohydrates. The β coefficients were similar but the relationship with polysaccharides only was statistically significant. Conversely to carbohydrates, a positive and significant association was observed between fat and BMI (in absolute term or in proportion of energy). This relationship became statistically not significant when carbohydrates and fat were analyzed together. These data were in agreement with the majority of epidemiological studies showing a negative relationship between obesity and carbohydrates and a positive one with fat.22,23,24,25 On the other hand, there were no statistically significant relationships between fat or carbohydrates and WHR when analyzed independently or when fat and carbohydrates were included in the same analysis. These different results between BMI and WHR suggest that diet composition may play a role in global fat accumulation but may not influence body fat distribution.26,27 It has been suggested that the increase of eating frequency may influence body weight regulation by reducing lipogenesis, and by reducing postprandial blood lipid levels and insulin secretion.28,29,30
In this population, men eating five times a day or more were, on average, leaner than those who ate once or twice a day, with BMI and WHR lower than 13 and 4% respectively.
Eating frequency accounted for 3 and 2.5% of total variability for BMI and WHR, respectively. These men's specificities differed from many other social and environmental characteristics. They were more educated, they smoked less and practised more leisure time physical activity. Diet composition and respective nutrient intakes varied greatly between these groups. In the high-eating-frequency group, carbohydrate consumption (as a percentage of energy) was 19% higher and fat and protein intakes about 9 and 15% lower, respectively, than in the low-eating-frequency group. The results emphasized the complexity of several factors and their interactions in the regulation of body weight. In addition, the level of the coefficients of multiple determination observed in the multivariate analysis (around 20%) suggests that a great number of parameters like genetic, environmental and psycho-social factors, not taken into account in this study, may play a determinant role in body fatness.
Maillard G, Charles MA, Thibult N, Forhan A, Sermet C, Basdevant A, Eschwege E . Trends in the prevalence of obesity in the French adult population between 1980 and 1991 Int J Obes Relat Metab Disord 1999 23: 389–394.
Flegal KM, Carroll MD, Kuczmarski RJ, Johnson CL . Overweight and obesity in the United States: prevalence and trends, 1960–1994 Int J Obes Relat Metab Disord 1998 22: 39–47.
Pietinen P, Vartiainen E, Mannisto S . Trends in body mass index and obesity among adults in Finland from 1972 to 1992 Int J Obes Relat Metab Disord 1996 20: 114–120.
Kuulasmaa K, Tunstall-Pedoe H, Dobson A, Fortmann S, Sans S, Tolonen H, Evans A, Ferrario M, Tuomilehto J . Estimation of contribution of changes in classic risk factors to trends in coronary-event rates across the WHO MONICA Project populations Lancet 2000 355: 675–687.
Bellisle F, McDevitt R, Prentice AM . Meal frequency and energy balance Br J Nutr 1997 77 (Suppl 1): S57–70.
Gatenby SJ . Eating frequency: methodological and dietary aspects Br J Nutr 1997 77 (Suppl 1): S7–20.
Edelstein SL, Barrett-Connor EL, Wingard DL, Cohn BA . Increased meal frequency associated with decreased cholesterol concentrations; Rancho Bernardo, CA, 1984–1987 Am J Clin Nutr 1992 55: 664–669.
Kant AK, Schatzkin A, Graubard BI, Ballard-Barbash R . Frequency of eating occasions and weight change in the NHANES I Epidemiologic Follow-up Study Int J Obes Relat Metab Disord 1995 19: 468–474.
Drummond SE, Crombie NE, Cursiter MC, Kirk TR . Evidence that eating frequency is inversely related to body weight status in male, but not female, non-obese adults reporting valid dietary intakes Int J Obes Relat Metab Disord 1998 22: 105–112.
Titan SM, Bingham S, Welch A, Luben R, Oakes S, Day N, Khaw KT . Frequency of eating and concentrations of serum cholesterol in the Norfolk population of the European prospective investigation into cancer (EPIC-Norfolk): cross sectional study Br Med J 2001 323: 1286–1288.
Evans AE, Ruidavets JB, McCrum EE, Cambou JP, McClean R, Douste-Blazy P, McMaster D, Bingham A, Patterson CC, Richard JL, Mathewson ZM, Cambien F . Autres pays, autres coeurs? Dietary patterns, risk factors and ischaemic heart disease in Belfast and Toulouse Q J Med 1995 88: 469–477.
Renaud S, Attier NC . La composition des aliments INSERM: Paris 1986.
Répertoire général des aliments. Table de composition REGAL Inra/Tec et Doc: Paris 1997.
Goldberg GR, Black AE, Jebb SA, Cole TJ, Murgatroyd PR, Coward WA, Prentice AM . Critical evaluation of energy intake data using fundamental principles of energy physiology: 1. Derivation of cut-off limits to identify under-recording Eur J Clin Nutr 1991 45: 569–581.
Schofield WN . Predicting basal metabolic rate, new standards and review of previous work Hum Nutr Clin Nutr 1985 39 (Suppl 1): 5–41.
SAS Institute Inc. SAS/STAT user's guide: the GLM procedure. Release 6.03 edn SAS Institute Inc.: Cary, NC 1990.
Marques-Vidal P, Ruidavets JB, Cambou JP, MD, Ferrières J . Trends in overweight and obesity in middle-aged subjects from Southwestern France, 1985–1997 Int J Obes Relat Metab Disord 2002 26: 732–734.
Black AE, Goldberg GR, Jebb SA, Livingstone MB, Cole TJ, Prentice AM . Critical evaluation of energy intake data using fundamental principles of energy physiology: 2. Evaluating the results of published surveys Eur J Clin Nutr 1991 45: 583–599.
Rogers PJ . Interrelationships between feeding patterns and obesity in humans Proceedings of the 1st European Congress on Obesity Stockholm 1988.
Summerbell CD, Moody RC, Shanks J, Stock MJ, Geissler C . Relationship between feeding pattern and body mass index in 220 free-living people in four age groups Eur J Clin Nutr 1996 50: 513–519.
Barrett-Connor E . Nutrition epidemiology: how do we know what they ate? Am J Clin Nutr 1991 54 (1 Suppl): 182S–187S.
Nelson LH, Tucker LA . Diet composition related to body fat in a multivariate study of 203 men J Am Diet Assoc 1996 96: 771–777.
Hill JO, Melanson EL, Wyatt HT . Dietary fat intake and regulation of energy balance: implications for obesity J Nutr 2000 130 (2S Suppl): 284S–288S.
Bolton-Smith C, Woodward M . Dietary composition and fat to sugar ratios in relation to obesity Int J Obes Relat Metab Disord 1994 18: 820–828.
Bray GA, Popkin BM . Dietary fat intake does affect obesity! Am J Clin Nutr 1998 68: 1157–1173.
George V, Tremblay A, Despres JP, Leblanc C, Bouchard C . Effect of dietary fat content on total and regional adiposity in men and women Int J Obes 1990 14: 1085–1094.
Larson DE, Hunter GR, Williams MJ, Kekes-Szabo T, Nyikos I, Goran MI . Dietary fat in relation to body fat and intraabdominal adipose tissue: a cross-sectional analysis Am J Clin Nutr 1996 64: 677–684.
Fabry P, Tepperman J . Meal frequency—a possible factor in human pathology Am J Clin Nutr 1970 23: 1059–1068.
Jenkins DJ, Wolever TM, Vuksan V, Brighenti F, Cunnane SC, Rao AV, Jenkins AL, Buckley G, Patten R, Singer W et al. Nibbling versus gorging: metabolic advantages of increased meal frequency New Engl J Med 1989 321: 929–934.
Juhel C, Pafumi Y, Senft M, Lafont H, Lairon D . Chronically gorging v. nibbling fat and cholesterol increases postprandial lipaemia and atheroma deposition in the New Zealand white rabbit Br J Nutr 2000 83: 549–559.
About this article
Cite this article
Ruidavets, J., Bongard, V., Bataille, V. et al. Eating frequency and body fatness in middle-aged men. Int J Obes 26, 1476–1483 (2002). https://doi.org/10.1038/sj.ijo.0802143
- eating frequency
- body fatness
- body mass index
- waist-to-hip ratio
The role of energy intake and energy misreporting in the associations between eating patterns and adiposity
European Journal of Clinical Nutrition (2018)
Nutritional quality of meals and snacks assessed by the Food Standards Agency nutrient profiling system in relation to overall diet quality, body mass index, and waist circumference in British adults
Nutrition Journal (2017)
Eating frequency is inversely associated with blood pressure and hypertension in Korean adults: analysis of the Third Korean National Health and Nutrition Examination Survey
European Journal of Clinical Nutrition (2014)
International Journal of Obesity (2014)
Effects of Manipulating Eating Frequency During a Behavioral Weight Loss Intervention: A Pilot Randomized Controlled Trial