To study the effect to recommend no snacks vs three snacks per day on 1-year weight loss. The hypothesis was that it is easier to control energy intake and lose weight if snacks in between meals are omitted.
In total 140 patients (36 men, 104 women), aged 18–60 years and body mass index>30 kg/m2 were randomized and 93 patients (27 men, 66 women) completed the study. A 1-year randomized intervention trial was conducted with two treatment arms with different eating frequencies; 3 meals/day (3M) or 3 meals and 3 snacks/day (3+3M). The patients received regular and individualized counseling by dieticians. Information on eating patterns, dietary intake, weight and metabolic variables was collected at baseline and after 1 year.
Over 1 year the 3M group reported a decrease in the number of snacks whereas the 3+3M group reported an increase (−1.1 vs +0.4 snacks/day, respectively, P<0.0001). Both groups decreased energy intake and E% (energy percent) fat and increased E% protein and fiber intake but there was no differences between the groups. Both groups lost weight, but there was no significant difference in weight loss after 1 year of treatment (3M vs 3+3M=−4.1±6.1 vs −5.9±9.4 kg; P=0.31). Changes in metabolic variables did not differ between the groups, except for high-density lipoprotein that increased in the 3M group but not in 3+3M group (P<0.033 for group difference).
Recommending snacks or not between meals does not influence 1-year weight loss.
Although the importance of regular mealtimes is consistently advocated in obesity treatment (Wing et al., 1996; NIH, 1998; DPP, 2002; SBU, 2002; Wadden and Stunkard, 2002; Elfhag and Rossner, 2005), the role of eating frequency in obesity is unclear (WHO, 2003) and there is no clear evidence of the impact of in-between meal snacking and frequent eating patterns on energy balance and weight loss (Drummond et al., 1996; Kirk, 2000; Booth et al., 2004; Jebb, 2005).
Yet, snacking may play a role in obesity management as snacking may influence energy intake and thus body weight. Spreading the energy load over the day by including several snacks may reduce appetite (Speechly et al., 1999) and as a consequence, decrease energy intake and body weight. On the other hand, snacking between meals may not be satiating (Booth, 1988; Marmonier et al., 2002) and therefore, habitual snacking may be a factor driving energy intake up and instead increase body weight. Although, short experimental studies in obese subjects have not shown any effect of eating frequency on weight loss in energy-restricted diets (Garrow et al., 1981; Verboeket-van de Venne and Westerterp, 1993), epidemiological studies suggest a link between snacking and weight gain (Basdevant et al., 1993; Coakley et al., 1998; Levitsky et al., 2004). Subjects who regain weight after successful weight reduction also report more snacks than those who maintain their weight loss (Kayman et al., 1990). These findings are in line with a number of studies showing that a high meal frequency and snacking are related to a high energy intake (Dwyer et al., 2001; Zizza et al., 2001; Bertéus Forslund et al., 2002, 2005). The effect of eating frequency is important to understand and an evidence-based appraisal is needed (Bellisle et al., 1997; de Graaf, 2000; Bray and Bouchard, 2004; Mattson, 2005; Parks and McCrory, 2005). If snacking increases the total energy intake the recommendation to eat snacks in between meals may be questioned in obesity treatment. For that reason longer, randomized interventions in free-living obese subjects are needed to elucidate the role of snacking in obesity treatment. To our knowledge no such studies exist.
The aim of this study was therefore to study the effect of two different recommended eating frequencies on 1-year weight loss in a randomized design. The hypothesis was that it is easier to control energy intake and lose weight if food intake is concentrated to three main meals per day compared to three main meals and three snacks.
A 1-year, parallel group, randomized clinical trial was conducted with two treatment groups with different eating frequencies; three meals per day (3M) or three meals and three snacks per day (3+3M). The study was conducted at three medical outpatient clinics in the western and southern part of Sweden (Sahlgrenska Hospital, Skaraborg Hospital and Helsingborg Hospital). The recruitment period was from September 2002 to January 2005 and the intervention period from September 2002 to January 2006. The study was coordinated from the obesity unit at Sahlgrenska University Hospital, Göteborg and at each study site a local dietician, physician and nurse were responsible for the running of the study. All participants received written and oral information about the study protocol from the registered dietician at each site and gave written informed consent. The study was approved by the ethics committees at the Faculty of Medicine, Göteborg University (Göteborg and Skövde) and Faculty of Medicine, Lund University (Helsingborg).
Body weight, height (only at baseline), waist and hip circumference and blood pressure were measured and fasting blood samples were collected at baseline and after 1 year. In addition, body weight was measured at every visit. Self-administered questionnaires including information on eating frequency, energy intake and physical activity were also completed at baseline and after 1 year. The primary outcome was change in weight after 1 year of treatment. Secondary outcomes included changes in blood pressure, cardiovascular risk factors, energy intake, eating frequency and the subjects' own evaluation of the treatment.
The study was a 1-year intervention with dietician counseling at a regular basis. Before study start, all patients met a dietician at a screen visit and received written and oral information about the study. From start of the study to the inclusion visit the patient met the dietician every 2 weeks up to week 12 and thereafter, every 4 weeks up to week 52. In total, 17 visits were offered, from inclusion visit to week 52. Each visit lasted for approximately 45 min.
At the inclusion visit the dietician gave instruction about the allocated eating frequency. As a guide for portion sizes and meal/snack composition an individualized energy-restricted, nutritionally balanced diet plan was prepared and handed out to the patients at the next visit. The calculations of prescribed energy level were based on basal metabolic rate (BMR) estimated according to the formula of Harris and Benedict (1919). From BMR, total daily energy expenditure was calculated by multiplying a physical activity level (PAL) × 1.3 for moderate physical activity and PAL × 1.5 for heavy physical activity. From the estimated total energy expenditure 30% was subtracted to get the prescribed energy intake. The minimum energy level prescribed was 1400 kcal/day. The prescribed energy level was divided into three meals or three meals and three snacks depending on which group the patient was randomized to. Recommended energy intake in the group of 3M was divided in breakfast, 30% of daily energy intake (D%), lunch 35D% and dinner 35D% and no snacks with the exception of limited fruit intake and calorie-free drinks. For the group of 3+3M the daily energy intake was divided in breakfast 20D%, lunch 25D%, dinner 25D% and three snacks, each on 10D%. In all other respects the prescribed diet followed Swedish Nutrition Recommendations (SNR) (Livsmedelsverket, 1997). The patients were encouraged to follow the allocated eating frequency throughout the study and the individualized diet plan was used as a guideline to enable changing eating behavior. In addition, the patients were encouraged to increase their physical activity, primarily walking on a regular basis.
Thus, the patients received individual counseling in changes of diet and physical activity behavior. A diet-counseling plan was followed by the dieticians to ensure a concordant treatment between the study sites. The diet-counseling plan included themes for every visit, nutritional information, fact sheets and self-monitoring exercises. Food and physical activity records could be used as a pedagogic tool. Even if each visit had a preplanned topic the counseling was individualized, focusing on specific individual problems. However, adherence to the allocated eating pattern was emphasized at all visits.
Compliance to the recommended eating frequency was evaluated by repeated telephone interviews at six predefined time periods during the year of intervention. The interviews were carried out by the dietician, who coordinated the study at Sahlgrenska University Hospital. ‘The meal pattern questionnaire’ was used as a basis for the assessment of intake occasions (Bertéus Forslund et al., 2002). The subjects were asked about their intake pattern the previous day specifying time and type of intake occasions. Food choices at snack meals were registered specifically; other intake occasions were registered according to the meal types in the questionnaire. The telephone interviews were conducted on randomly selected days with emphasis to cover different days of the week. If it was impossible to get in contact with the subject in the predefined time period the interview was omitted in this period.
Body weight was measured to the nearest 0.05 kg with the patient wearing underwear and no shoes, using calibrated scales. Body height was measured without shoes to the nearest 0.05 cm. Body mass index (BMI) was calculated from weight (kg) divided in height squared (m2). Waist circumference was measured in a standing position at the midpoint between lower border of the rib cage and the iliac crest. Hip was measured at the symphysis major trochanter level.
Blood pressure and biochemical analyses
Blood pressure was measured after 5 min in a sitting position on the right arm. Blood samples; P-glucose, S-insulin, S-cholesterol, S-high-density lipoprotein (HDL), S-low-density lipoprotein (LDL) and S-triglycerides were drawn in a fasting state, that is no food or drink were allowed from 1200 the night before measurement day. Blood samples were analyzed locally at the central clinical laboratories at Sahlgrenska University Hospital, Skaraborg Hospital and Helsingborg Hospital. Laboratory analyses were the same as those used in ordinary patient care according to local practice.
Assessment of eating pattern
A self-administered questionnaire, ‘The meal pattern questionnaire’, was used to assess habitual daily intake pattern. The questionnaire was distributed at baseline and at the end of study. The subjects were asked to describe how they eat ‘an ordinary’ day, specifying time for each intake occasion and choose one of four predefined types of intake occasions; main meal, light meal/breakfast, snacks and drink only. In the analysis of the eating pattern, main meals and light meal/breakfast were added together and called principal meals (one light meal/breakfast and two main meals or two light meals/breakfast and one main meal). The questionnaire is described elsewhere (Bertéus Forslund et al., 2002).
Assessment of dietary intake
A self-administered dietary questionnaire to assess habitual energy and macronutrient intake during the past 3 months was used. The questionnaire was distributed at baseline and at the end of the study. The dietary questionnaire is judged to give valid results in both obese and normal weight subjects. The questionnaire is described elsewhere (Lindroos et al., 1993).
Assessment of physical activity
A questionnaire describing physical activity at work and during leisure time was used (Larsson et al., 2004). Occupational PAL was categorized in five levels; unemployed, sedentary work, moderately sedentary work, moderately heavy work and heavy work. Leisure time physical activity was categorized in four levels; sedentary leisure, moderately activity, moderate exercise and heavy exercise. The participants choose one of the alternatives corresponding to their usual activity pattern. In our analyses, the leisure time activity level ‘sedentary leisure’ and occupational PAL ‘sedentary work’ were coded as sedentary in a dichotomous variable, sedentary yes=1, no=0.
To evaluate the subjects' own opinion on the allocated eating pattern (3M or 3+3M) they were asked to answer the questions on a Visual Analog Scale: ‘How content are you with eating 3 (3+3) meals per day?’ (not content=0, very content=100). ‘How easy did you find it eating 3 (3+3) meals per day? (very difficult=0, very easy=100). ‘Would you consider eating 3 (3+3) meals per day from now on?’ (Yes=1/No=2).
Patients referred to the obesity unit at Sahlgrenska University Hospital, Göteborg, obesity research unit at Helsingborg Hospital, Helsingborg and at the Medical clinic at Skaraborgs Hospital, Skövde were invited to participate at the first visit to the clinics. At the latter, clinic participants were also recruited through local advertisement. The patients were recruited continuously over time, starting at Sahlgrenska University Hospital in September 2002. To speed up recruitment Helsingborg Hospital joined in March 2003 and Skaraborgs Hospital in March 2004.
The selection criteria to enter the study included age 18–60 years and BMI>30 kg/m2. Subjects reporting previous obesity surgery, anti-obesity drug treatment the last year, drug- or insulin-treated diabetes, hypothyroidism, severe psychiatric disorder, bulimia, drug or alcohol abuse were not eligible for the study.
Pre-study power calculations showed that 70 subjects were needed in each group to obtain a significant (P<0.05) difference in body weight change of 3.0±5.2 kg with a power of 80% and an estimated dropout rate of 35%. Accordingly, two groups of each 70 patients were randomly allocated to the two different intervention groups; three meals or three meals and three snacks per day. A block randomization was used to keep the two groups balanced at all times and evenly spread throughout the year, according to Altman (1991). The two groups were in blocks of four at a time. In each block two subjects got group ‘three meals’ and two subjects got group ‘three meals+three snacks per day’ in a random order. Blinded and sealed envelopes for the randomization were prepared at the Sahlgrenska site and sent out to the two other sites. The procedure was supervised from the Sahlgrenska University Hospital and the sites were in contact continuously. Randomization was carried out at the inclusion visit and the dietician gave instruction about the allocated eating frequency.
To analyze differences between groups χ2 test was used for proportions, McNemars’ test for paired proportions and t-test for continuous variables. Survival analysis was used to compare time for dropout in the two study groups. Weight, anthropometry and laboratory variables were analyzed in completers and in all participants using the last observation carried forward (LOCF). Repeated measures analysis was used to analysis weight change between study groups over time. The SAS 8.2 statistical package was used for all analyses (SAS Institute Inc., Cary, NC, USA).
A total of 140 (36 men and 104 women) patients were randomized and 93 (27 men and 66 women) patients completed the entire study. Participation flow is shown in Figure 1. Dropout rate was 30% in the 3M group and 37.1% in the 3+3M group, although the difference was not statistical significant (P=0.37). There was no difference in time of attrition between the study groups (P=0.27). However, younger patients (P=0.004) and patients with lower BMI (P=0.01) dropped out from the study program earlier than older patients and those with higher BMI. In addition more men in the 3m group dropped out compared to the 3+3M group (7/18 compared to 2/18, respectively; P=0.05) whereas in women dropout rate was higher in the 3+3M group than the 3M group (24/52 and 14/52, respectively; P=0.04). Baseline characteristics for all study participants and for completers in both groups are shown in Table 1. Baseline characteristics did not differ significantly between completers and all participants included in the study. Neither did the participants who completed the study differ between the three study sites.
Intake of eating occasions at baseline and after 1 year of treatment is presented in Table 2. The change in number of principal meals per day did not differ between the 3M and 3+3M group. However, change in number of snacks differed significantly between the two groups. The 3M group decreased the number of snacks whereas the 3+3M group increased snack frequency (P<0.0001, confidence interval (CI) −2.18 to −1.06). Figures 2a and b show the percent completers in each group reporting number of principal meals and snacks before treatment and after 1 year.
After 1 year of treatment 22 patients (45%) in the 3M group reported consuming the recommended three principal meals and no snacks whereas 21 patients (48%) in the 3+3M group reported having the recommended three principal meals and three snacks.
Diet and physical activity
Mean energy and macronutrient intake and physical activity at baseline and after 1 year of treatment is shown in Table 2. Reported energy intake decreased with 2955 kJ (707 kcal) in the 3M group compared to 2178 kJ (521 kcal) in the 3+3M group and the decrease did not differ significantly between the two groups. The reported change in energy intake was in men 4140 kJ (991 kcal) and 2021 kJ (484 kcal) in 3 and 3+3M groups, respectively. Corresponding figures for women were 2584 kJ (618 kcal) and 2274 kJ (544 kcal) in the 3 and 3+3M groups, respectively. Furthermore, change in energy percent macronutrient intake did not differ between the two groups. Although both groups decreased the energy percent fat intake and increased energy percent protein and fiber intake expressed as g/1000 kcal from baseline to week 52.
After 1 year of treatment number of patients reporting sedentary lifestyle decreased significantly in both groups and there was no significant difference between the groups. Neither did changes in sedentary work differ between the groups (Table 2).
Repeated interviews on eating frequencies with emphasis on snacking were conducted throughout the study. Mean number of interviews was 4.4 per subject. Reported mean number of principal meals and snacks is described in Table 3. The 3M group reported fewer snacks than the 3+3M group. In the 3+3M group the frequency of snacks was decreasing in the latter study period. The results of compliance are in line with the meal frequency reported by the subjects at the end of study as described in the section ‘Eating frequency’.
Weight loss after 1 year of treatment was in the 3M group −4.1±6.1 kg (−3.6±4.9%) and in the 3+3M group −5.9±9.4 kg (−4.7±6.7%) and did not differ significantly (P=0.31). When analyzing weight loss over time no difference was found between the two groups neither in the completers only (P=0.34) nor in all participants using LOCF (P=0.35) (Figures 3a and b).
Changes in blood pressure, blood glucose, insulin, cholesterol, LDL, HDL and triglycerides did not differ between the groups. However, HDL increased in the 3M group compared to the 3+3M group (P<0.033) (Table 4).
The patients' personal opinion on the meal regimen was evaluated. When analyzing the question ‘How content are you with eating 3 (3+3) meals per day?’ no difference was found between the two groups. The mean score was 55 and 63% in the 3M and 3+3M groups (P=0.14), respectively. Neither was a difference found between the groups replying the question ‘How easy did you find it eating 3 (3+3) meals per day?’ showing a mean score of 50 and 55% in the 3M and 3+3M group (P<0.30), respectively. Nor was a difference found when asking ‘Would you consider eating 3 (3+3) meals per day from now on?’; 51% of the patients in the 3M group reported ‘yes’ compared to 68% in the 3+3M group (P<0.10).
In this 1-year randomized clinical trial subjects in both groups lost weight and improved their metabolic profile over 1 year. However, weight loss did not differ significantly between the two intervention arms suggesting that recommending snacks or not between meals is not an important factor for achieved weight loss after 1 year. As previous cross-sectional studies have shown that a high eating frequency and snacking increase total energy intake (Bertéus Forslund et al., 2002, 2005), we hypothesized that no snacking between meals would facilitate the restriction of energy intake and weight loss. Even if omitting snacks may help cutting down energy intake, our result implies that when patients attain extensive support and diet counseling they manage to cut down calories despite a high snacking frequency. The choice of low-energy dense snacks is crucial and we can only speculate if the good quality snack choices will be sustained without extensive support. It may be suggested that the role of snacking is different in obese ‘real life’ and during treatment conditions. Not only frequency but regularity of meal times may also have an impact on energy intake. In a recent study by Farshchi et al. (2005) obese women were instructed to maintain their usual intake on an irregular (‘caotic’ pattern with 3–9 meals/day) vs an regular (6 meals/day) meal pattern in a 14-day crossover design. The obese women reported a significantly higher energy intake during the irregular meal pattern than during the regular meal pattern. In a similar study in lean women, energy intake did not differ between the two meal patterns suggesting that eating patterns may have different implications in normal weight and obese subjects (Farshchi et al., 2004). Although we do not know how regular the patients were eating during the intervention, it is possible that the extensive support helped the patients to follow a more regular meal pattern.
A crucial point when evaluating our weight loss results is the adherence to the allocated intervention. Both groups changed eating patterns toward the recommended number of snacks and the reported number of snacks differed significantly between the groups after 1 year. Adherence was also similar in the two groups. This suggests that many subjects in the present study managed to change eating patterns despite the difficulties in doing so reported by other investigators (King and Gibney, 1999). It is noteworthy that the subject's own opinion on difficulties did not differ between the groups. However it should be noted that the discrepancy in snacking between the two groups was not as large as intended. This suggests that the difference in snacking patterns might not have been large enough to attain a difference in weight loss.
The weight loss difference between the treatment arms was 1.8 kg. It may be argued that we did not have enough statistical power to find a difference due to too small study groups. When planning the study we decided that a difference of 3 kg or more would be considered clinically relevant in a weight loss trial. This is in line with anticipated weight loss differences used in power calculations in other studies (Heshka et al., 2003; Samaha et al., 2003; Brinkworth et al., 2004).
Previous studies on eating patterns have focused mainly on the influence on metabolic factors. Spreading the nutrient load on many small meals may reduce insulin and glucose response and improve blood lipid profile (Fábry et al., 1964; Jenkins et al., 1989, 1992) although findings are inconsistent (Beebe et al., 1990; Arnold et al., 1994, 1997; Thomsen et al., 1997). In this study metabolic variables were improved in both groups after 1 year but did not differ except for HDL cholesterol that increased in the 3M group. The literature on eating frequency and HDL cholesterol is inconsistent. In short experimental studies HDL cholesterol has been positively (McGrath and Gibney, 1994), negatively (Murphy et al., 1996; Thomsen et al., 1997) or unrelated (Arnold et al., 1993, 1994) to eating frequency. Therefore, we cannot role out that the difference in HDL cholesterol is a chance finding.
Attrition is usually high in obesity treatment studies (Glenny et al., 1997). The dropout rate in this study was similar to what we had expected and in line with withdrawals found in other studies (Clark et al., 1995; Torgerson et al., 1999). In line with other studies we also found that younger patients dropped out earlier than older patients (Andersson and Rossner, 1997; Torgerson et al., 1999; Lantz et al., 2003a, 2003b). However, patients with lower BMI dropped out earlier, which is in contrast to others that found no association between BMI and attrition (Andersson and Rossner, 1997; Torgerson et al., 1999; Lantz et al., 2003a) or that those with higher BMI dropped out more frequently (Clark et al., 1995). One study with a very high dropout rate (77%) also found that dropouts had a slightly lower BMI than completers (Inelmen et al., 2005).
The larger withdrawal in men from the 3M group and women from the 3+3M group indicates that preferred snacking frequency may differ by gender. We can only speculate if men find it easier to adhere to a frequent snacking pattern than no snacks whereas women do the opposite. Although, gender differences have also been noted in a previous intervention study. This study showed that men who adhered to three principal meals and two or three snacks per day lost more weight than those who did not, whereas women who adhered to this eating pattern lost less weight than those who did not (H Bertéus Forslund, personal communication). Gender differences have also been noted in observational studies suggesting a negative association between meal frequency and BMI or body weight in men and a positive or no relationship in women (Drummond et al., 1998; Titan et al., 2001).
Thus, one limitation of the present study is that we lack power to analyze gender differences. Another limitation is that the recruitment period was very long. To speed up recruitment we involved two other study sites. This made the study more heterogeneous. On the other hand recruiting subjects from different parts of Sweden strengthens the generalizablity of the results.
In Sweden, the commonly used dietary recommendation in obesity treatment is based on the general dietary recommendations for the whole Swedish population, SNR (Livsmedelsverket, 1997). The SNR recommendations include eating frequency as well as temporal distribution of energy over the day. An eating frequency of 3 main meals and 2–3 snacks has been recommended, although revised to 1–3 snacks/day recently (Livsmedelsverket, 2005). However, the evidence that this recommendation facilitates energy restriction and weight control is not substantiated. This study showed that approximately half of the patients in each group managed to adhere to the allocated ‘no snack’ or ‘three-snack’ pattern and also considered to continue this eating pattern after the study had ended. The findings from this study also suggest that a recommended energy-restricted diet results in similar weight loss irrespective high- or low-eating frequency. Consequently, recommending snacks in obesity treatment should be based on individual needs rather than that all patients should eat snacks or not.
We thank Ted Lystig for statistical advice. This study was supported by a grant from Västra Götalandsregionen.
About this article
Nutrition Journal (2012)