Diabetologia (2014) 57:1552–1560

DOI 10.1007/s00125-014-3253-5

ARTICLE

Eating two larger meals a day (breakfast and lunch) is more

effective than six smaller meals in a reduced-energy regimen
for patients with type 2 diabetes: a randomised crossover study
Hana Kahleova & Lenka Belinova & Hana Malinska & Olena Oliyarnyk &
Jaroslava Trnovska & Vojtech Skop & Ludmila Kazdova & Monika Dezortova &
Milan Hajek & Andrea Tura & Martin Hill & Terezie Pelikanova

Received: 22 January 2014 / Accepted: 9 April 2014 / Published online: 18 May 2014
# The Author(s) 2014. This article is published with open access at Springerlink.com

Abstract −2,092 kJ. The diet in both regimens had the same macronu-
Aims/hypothesis The aim of the study was to compare the trient and energy content. HFC was measured by proton
effect of six (A6 regimen) vs two meals a day, breakfast and magnetic resonance spectroscopy. Insulin sensitivity was
lunch (B2 regimen), on body weight, hepatic fat content measured by isoglycaemic–hyperinsulinaemic clamp and cal-
(HFC), insulin resistance and beta cell function. culated by mathematical modelling as oral glucose insulin
Methods In a randomised, open, crossover, single-centre sensitivity (OGIS). Beta cell function was assessed during
study (conducted in Prague, Czech Republic), we assigned standard meal tests by C-peptide deconvolution and was
54 patients with type 2 diabetes treated with oral quantified with a mathematical model. For statistical analysis,
hypoglycaemic agents, both men and women, age 30–70 2×2 crossover ANOVA was used.
y e a r s , B M I 2 7 – 5 0 k g / m 2 a n d H b A 1 c 6 – 11 . 8 % Results The intention-to-treat analysis included all participants
(42–105 mmol/mol), to follow two regimens of a (n=54). Body weight decreased in both regimens ( p<0.001),
hypoenergetic diet, A6 and B2, each for 12 weeks. more for B2 (−2.3 kg; 95% CI −2.7, −2.0 kg for A6 vs −3.7 kg;
Randomisation and allocation to trial groups (n=27 and 95% CI −4.1, −3.4 kg for B2; p<0.001). HFC decreased in
n=27) were carried out by a central computer system. Indi- response to both regimens ( p<0.001), more for B2 (−0.03%;
vidual calculations of energy requirements for both regimens 95% CI −0.033%, −0.027% for A6 vs −0.04%; 95% CI
were based on the formula: (resting energy expenditure×1.5) −0.041%, −0.035% for B2; p=0.009). Fasting plasma glucose
and C-peptide levels decreased in both regimens ( p<0.001),
H. Kahleova (*) : L. Belinova : H. Malinska : O. Oliyarnyk : more for B2 ( p=0.004 and p=0.04, respectively). Fasting plas-
J. Trnovska : V. Skop : L. Kazdova : T. Pelikanova ma glucagon decreased with the B2 regimen ( p<0.001), where-
Diabetes Centre, Institute for Clinical and Experimental Medicine,
Videnska 1958/9, 140 21 Prague, Czech Republic
as it increased ( p=0.04) for the A6 regimen ( p<0.001). OGIS
e-mail: hana.kahleova@gmail.com increased in both regimens ( p<0.01), more for B2 ( p=0.01).
No adverse events were observed for either regimen.
L. Belinova Conclusions/interpretation Eating only breakfast and lunch
First Faculty of Medicine, Charles University,
reduced body weight, HFC, fasting plasma glucose, C-peptide
Prague, Czech Republic
and glucagon, and increased OGIS, more than the same caloric
M. Dezortova : M. Hajek restriction split into six meals. These results suggest that, for
Department of Diagnostic and Interventional Radiology, Institute for type 2 diabetic patients on a hypoenergetic diet, eating larger
Clinical and Experimental Medicine, Prague, Czech Republic
breakfasts and lunches may be more beneficial than six smaller
A. Tura meals during the day.
Metabolic Unit, Institute of Biomedical Engineering, National Trial registration ClinicalTrials.gov number, NCT01277471,
Research Council, Padua, Italy completed.
Funding Grant NT/11238-4 from Ministry of Health, Prague,
M. Hill
Department of Steroid Hormones and Proteohormones, Institute of Czech Republic and the Agency of Charles University – GAUK
Endocrinology, Prague, Czech Republic No 702312.
Diabetologia (2014) 57:1552–1560 1553

Keywords Hepatic fat content . Insulin sensitivity . Two been observed that eating breakfast regularly may protect
meals a day . Type 2 diabetes against weight gain, despite a higher total daily energy intake
[14].
Abbreviations To the best of our knowledge, no interventional trials have
A6 Six meals a day regimen investigated the relationship between eating frequency and
B2 Two meals a day regimen weight change together with hepatic fat content (HFC), glu-
HFC Hepatic fat content cose tolerance and insulin resistance in humans, especially in
MCR Metabolic clearance rate of glucose patients with type 2 diabetes. The aim of our study was to
OGIS Oral glucose insulin sensitivity compare the effect of six vs two meals a day (breakfast and
REE Resting energy expenditure lunch, as this regimen allows a reasonable fasting time, yet is
sustainable in the long term) with the same caloric restriction
on body weight, HFC, insulin resistance and beta cell function
in individuals with type 2 diabetes. It was hypothesised that
Introduction eating only breakfast and lunch would reduce body weight
and HFC (and consequently, improve insulin resistance and
Frequency of meals is an important aspect of nutrition, with beta cell function) more than six meals a day would.
profound effects on human health and lifespan. Excessive ener-
gy intake is associated with an increased incidence of chronic
diseases including diabetes and is a leading cause of disability Methods
and death in Western countries [1]. A hypoenergetic diet is
crucial for both the prevention and treatment of type 2 diabetes. Participants
It is usually consumed as five or six small meals per day. Eating
more frequently is presumed to reduce hunger and thus reduce Out of the 219 individuals screened, 54 patients with
energy intake and body weight. However, the effects of meal type 2 diabetes (with disease duration of more than
frequency on human health and longevity are unclear [2]. 1 year) treated by oral hypoglycaemic agents (both men
Reduced meal frequency can prevent the development of and women), age 30–70 years, BMI 27–50 kg/m2 and HbA1c
chronic diseases and extend the lifespan in laboratory animals 6–11.8% (42–105 mmol/mol), met all the inclusion criteria,
due to lower oxidative damage and higher stress resistance [3, gave their written informed consent and underwent
4]. Mice under time-restricted feeding have an equivalent randomisation. Exclusion criteria comprised alcohol or drug
energy intake from a high-fat diet as those with ad libitum abuse, pregnancy or lactation, unstable medication or weight
access yet are protected against obesity, hyperinsulinaemia and in the last 3 months, a diagnosis of type 1 diabetes and the
hepatic steatosis [5, 6]. Intermittent fasting leads to a prolonged presence of a cardiostimulant.
lifespan and positively affects glucose tolerance, insulin sensi-
tivity and incidence of type 2 diabetes in mice [3, 4]. There is Study design
also emerging literature demonstrating a relationship between
the timing of feeding and weight regulation in animals. We used a randomised crossover study design. The study
Observational trials in humans indicate that eating more protocol was approved by the Institutional Ethical Committee.
often than three times a day may play a role in overweight and In a single-centre study, after a 1 month run-in period (when
obesity [7] and that frequent eating predisposes to a higher the patients learned how to write their food diaries and use the
energy intake by increasing food stimuli and difficulty con- pedometers and glucometers), the participants began a
trolling energy balance [8]. In a randomised controlled study, 12 week regimen of either six (A6) or two (B2) meals a day.
more frequent eating was not related to a greater reduction in The A6 regimen consisted of three main meals (breakfast,
energy intake or body weight [9]. In type 2 diabetic patients it lunch and dinner), and three smaller snacks in between. The
has been demonstrated that it may be more beneficial for B2 regimen consisted of breakfast (eaten between 06:00 and
glycaemic control to eat one larger instead of two smaller 10:00 hours) and lunch (eaten between 12:00 and
meals, provided the diet is rich in fibre [10]. 16:00 hours). The regimens were switched for the subsequent
It has been demonstrated that a large isocaloric mixed meal 12 weeks. All measurements were performed at weeks 0
causes a greater postprandial thermogenic response than the (baseline), 12 and 24 (Fig. 1 and Table 1).
same food consumed in six smaller portions [11]. Observa-
tional data suggest that eating meals later in the day may Diet
influence the success of weight-loss therapy, even in humans
[12]. It has also been shown that fat storage increases during The composition of the diet in both regimens followed the
the day and is the greatest after an evening meal [13]. It has Study Group on Diabetes and Nutrition of the European
1554 Diabetologia (2014) 57:1552–1560

Fig. 1 Enrolment of the 219 persons with T2DM were

participants and completion of the December 2010–February 2011 screened
study. T2DM, type 2 diabetes
mellitus 165 were excluded
112 did not meet inclusion criteria
53 refused to participate

54 underwent
February 2011
randomisation

March 2011 27 started with 27 started with

6 meals a day 2 meals a day

2 withdrew owing to 1 withdrew owing to

lack of motivation personal reasons

25 completed 12 weeks 26 completed 12 weeks

June–July 2011 of 6 meals a day and of 2 meals a day and
started with started with
2 meals a day 6 meals a day

2 withdrew owing to 2 withdrew owing to

personal and family reasons lack of motivation

September– 23 completed 12 weeks 24 completed 12 weeks

October 2011 of 2 meals a day of 6 meals a day

27 were included in the 27 were included in the

analysis analysis

Association for the Study of Diabetes guidelines [15] with the country-specific food-nutrient database NutriDan 1.2 (www.
same caloric restriction: a restriction of 2,092 kJ/day institut-danone.cz/cz/odborna-sekce/nutridan).
(500 kcal/day) based on the measurement of each individual’s
resting energy expenditure (REE) by indirect calorimetry Physical activity
(metabolic monitor VMAX; SensorMedics, Anaheim, CA,
USA) [16]. Individual calculations of energy requirements This was assessed with an Omron HJ-720IT pedometer
for both regimens were based on the formula: (REE×1.5)− (Omron, Kyoto, Japan; using a 1 month average step count
2,092 kJ. The diet derived 50–55% of its total energy from for evaluation) and two questionnaires: the International
carbohydrates, 20–25% from protein and less than 30% from Physical Activity Questionnaire [17] and the Baecke ques-
fat (≤7% saturated fat, less than 200 mg/day of cholesterol), tionnaire [18] at weeks 0, 12, and 24.
with 30–40 g/day of fibre. Alcoholic beverages were limited
to one per day for women and two per day for men. Partici- Medication
pants were asked not to alter their exercise habits during the
study. Each regimen started with a 4 day tutorial where they Participants were asked to continue their pre-existing medica-
learned in detail how to compose and prepare their diet, with tion regimens, except when hypoglycaemia occurred repeat-
follow-up 1 h weekly meetings with lectures and cooking edly (fasting plasma glucose determined at the laboratory
classes throughout the whole study. All the meals during the <4.4 mmol/l or a capillary glucose reading <3.4 mmol/l
entire 24 weeks of the study were provided for one half of the accompanied by hypoglycaemic symptoms). In such cases,
participants (randomised within each study arm with an equal medications were reduced by a study physician following the
number of participants) while the other half of the participants medication protocol. All participants were given an Accu-Chek
prepared their meals by themselves. Performa glucometer (Roche, Basel, Switzerland) and
instructed how to use it.

Compliance Procedures

At weeks 0, 12, and 24, a 3 day dietary record (2 weekdays All measurements were performed on an outpatient basis at
and 1 weekend day) was completed by each participant. A weeks 0, 12 and 24, after a 10–12 h overnight fasting with tap
registered dietitian analysed all these dietary records using a water ad libitum. Height and weight were measured using a
Diabetologia (2014) 57:1552–1560 1555

Table 1 Baseline characteristics of the study population Hyperinsulinaemic isoglycaemic clamp The hyperinsulinaemic
Characteristic Study group (n=54) (1 mU kg−1 min−1) isoglycaemic clamp, lasting 3 h, was
conducted as previously described [19]. Insulin sensitivity
Age (years) 59.4±7.0 was estimated as the metabolic clearance rate of glucose
Sex, n (%) (MCR) [19].
Male 29 (54)
Female 25 (46) Proton magnetic resonance spectroscopy HFC was measured
Duration of diabetes (years) 8.1±5.8 by proton magnetic resonance spectroscopy on a 3 T MR
Smokers, n (%) 10 (19) scanner (Magnetom Trio, Siemens, Erlangen, Germany) with
Weight (kg) 94.1±15.5 an eight-channel body array coil. This method has been validat-
BMI (kg/m−2) 32.6±4.9 ed at our institution [20]. The measurement protocol included
HbA1c (DCCT) (%) 7.2±3.3 conventional MRI using a localiser and HASTE sequence with
HbA1c (IFCC) (mmol/mol) 54.9±13.0 breath-holding in the coronal and transversal planes. Spectra
Systolic blood pressure (mmHg) 140±14 were obtained from three different segments of the right lobe
Diastolic blood pressure (mmHg) 85±8 of the liver—volume of interest, 30 ml each and evaluated using
Resting heart rate (beats/min) 71±9 the LCModel (www.s-provencher.com/pages/lcmodel.shtml)
Oral hypoglycaemic agents, n (%) and MestReC (Mestrelab Research, Santiago de Compostela,
Metformin 41 (76) Spain) programs. The signal intensities of water and hepatic
Sulfonylurea 16 (30) lipids were used to determine the fat to total signal peak area
Thiazolidinedione 3 (6) ratio and then converted to absolute concentrations expressed as
Glinides 2 (4) a percentage of fat using equations validated by Longo et al [21].
Acarbose 1 (2) Fourteen individuals did not undergo an HFC measurement due
DPP-4 inhibitors 19 (35)
to the patient’s refusal, claustrophobia or the patient’s weight
Lipid-lowering therapy, n (%) 31 (57)
exceeding the limit of the equipment.
Antihypertensive therapy, n (%) 33 (61)
Calculations
Data are means ± SD
DCCT, Diabetes Control and Complications Trial; IFCC, International Modelling analysis of beta cell function was performed during
Federation of Clinical Chemistry standard meal tests. Insulin secretory rates were calculated
from plasma C-peptide levels by deconvolution [22] and
expressed per square meter of estimated body surface area.
periodically calibrated scale accurate to 0.1 kg. Waist circum- The dependence of insulin secretory rates on glucose levels
ference was measured with a tape measure placed at the was modelled separately for each patient and each study
midpoint between the lowest rib and the upper part of day. The beta cell model used in the present study, describing
the iliac bone. Blood pressure and heart rate were measured the relationship between insulin secretion and glucose
after 5 min in a seated position at rest, using a digital M6 concentration, has previously been described in detail
Comfort monitor (Omron, Kyoto, Japan). Three measure- [23–25].
ments were taken 2 min apart. The first measurement was Briefly, insulin secretion consists of two components. The
discarded, and the mean of the remaining two measurements first component represents the dependence of insulin secretion
was recorded. on absolute glucose concentration at any time point and is
characterised by a dose–response function. Characteristic var-
Indirect calorimetry Gas exchange measurements were taken iables of the dose–response are insulin secretion at a fixed
during a 45 min basal period before the clamp. Air flow and glucose concentration and the mean slope in the observed
O2 and CO2 concentrations in expired and inspired air were glucose range. The dose–response was modulated by a poten-
measured by a continuous open-circuit system (metabolic tiation factor that accounts for several agents (prolonged
monitor VMAX; SensorMedics, Anaheim, CA, USA). exposure to hyperglycaemia, non-glucose substrates, gastro-
intestinal hormones and neurotransmitters). The potentiation
Meal tests Plasma concentrations of glucose, immunoreactive factor was set to be a positive function of time and to be an
insulin and C-peptide were measured at 0, 30, 60, 120 and average of 1 during the experiment. It thus expresses a relative
180 min after a standard breakfast (1,895 kJ, 45% carbohy- potentiation of the secretory response to glucose.
drates, 17% proteins, 38% lipids). Insulin secretion and The second insulin secretion component represents a
whole-body insulin sensitivity were calculated by mathemat- dynamic dependence of insulin secretion on the rate of change
ical modelling (described below). of glucose concentration. Termed the derivative component, it
1556 Diabetologia (2014) 57:1552–1560

is described by a single variable, rate sensitivity. This secretion a b c

0 0 0
component is related to early insulin release [23, 24]. A6 B2 A6 B2 A6 B2

Δ Fasting plasma
glucose (mmol/l)
Δ Weight (kg)
-1 -0.01 -0.2

Δ HFC (%)
The model variables (the variables of the dose–response, -2 -0.02
-0.4
the rate sensitivity and the potentiation factor) were estimated -3 -0.03
-0.6
from the glucose and C-peptide concentrations by regularised -4 -0.04
*** -0.8
least squares, as previously described [23, 24]. Estimation of -5 -0.05 **
-1
the individual model variables was performed blinded for the **
randomisation of the patients for treatment. d e f
Whole-body insulin sensitivity was estimated in two ways: 0 200 0

Δ Fasting C-peptide (nmol/l)

Δ Fasting glucagon (pg/ml)

A6 B2 A6 B2
-0.05
(1) as the MCR calculated during the last 20 min of the 100

Δ HbA1c (IFCC, %)
-0.05
-0.1
isoglycaemic hyperinsulinaemic clamp after correction for 0
-0.15
-0.1 A6 B2
changes in glucose pool size [19], and (2) by a glucose–insulin -100
-0.2
model to derive an oral glucose insulin sensitivity (OGIS) -0.15 -200 -0.25

index, validated against the clamp data [3]. -0.2 * -300 -0.3
NS
-0.35
-400 ***
Analytical methods g h i
0.8 NS 30 0

Δ MCR (ml kg-1 min-1)

Δ OGIS (ml min-1 m-2)

*
Serum glucose was analysed using the Beckman Analyser 25 -100
A6 B2

Δ REE (kJ/day)
0.6
glucose-oxidase method (Beckman Instruments, Fullerton, 20 -200
0.4 15
CA, USA). Plasma immunoreactive insulin and C-peptide -300
0.2 10
-400
concentrations were determined using insulin and C-peptide 5
0 -500
IRMA kits (Immunotech, Prague, Czech Republic). HbA1c A6 B2 0
-600
A6 B2 NS
was measured by HPLC (Tosoh, Tokyo, Japan). Plasma con-
Fig. 2 Changes in anthropometric and laboratory variables. Data are
centrations of glucagon were measured using ELISA kits
shown as changes from baseline in response to the regimen of six (A6)
(BioVendor, Brno, Czech Republic). Plasma lipids concentra- and two meals (B2) a day. Data are mean ± 95% CI. Significance of the
tions were measured by enzymatic methods (Roche, Basel, factor treatment (assessed by 2×2 crossover ANOVA) is indicated by:
Switzerland). HDL-cholesterol was measured after double *p<0.05; **p<0.01; ***p<0.001; NS, non-significant. (a) Δ Weight,
n=54, (b) Δ HFC, n=48, (c) Δ Fasting plasma glucose, n=54, (d) Δ
precipitation with dextran and MgCl2. LDL-cholesterol was Fasting plasma C-peptide, n=54, (e) Δ Fasting plasma glucagon, n=54,
estimated using the Friedewald equation if the triacylglycerol (f) Δ HbA1c, n=54, (g) Δ MCR, n=49, (h) Δ Insulin sensitivity (OGIS),
concentration was <4.53 mmol/l. n=51, (i) Δ REE, n=52. To convert values for HbA1c in % into
mmol/mol, subtract 2.15 and multiply by 10.929
Statistical analyses

The intention-to-treat analysis included all participants. We regimens) were not significant. No substantial unfavourable
tested the distributions of the data. If the distribution was effects of the regimens were observed.
skewed, we used the Box-Cox transformation to attain data
symmetry and homoscedasticity [26]. Non-homogeneities in Dietary intake and physical activity
the data were detected using residual analysis as described
elsewhere [27]. 2×2 crossover ANOVA was used for data Reported dietary intake decreased (p<0.001) comparably un-
evaluation. The model consisted of the between-subject factor der both regimens. Physical activity increased ( p<0.05)
‘sequence’, the factor ‘subject’ and within-subject factors of slightly, but negligibly—by about 2,000 steps per month—in
‘period’ and ‘treatment’. In a subsequent subanalysis, the both regimens (see Table 2).
factor for prepared meals that were collected by patients was
added. The relationships between continuous variables were Body weight and HFC
evaluated using Pearson’s correlation and BMI-adjusted
partial correlations. Body weight and HFC decreased under both regimens
( p<0.001), more with B2 ( p<0.001; −2.3 kg; 95% CI −2.7,
−2.0 kg with A6 vs −3.7 kg; 95% CI −4.1, −3.4 kg with B2;
Results and p=0.009; −0.03%; 95% CI −0.033, −0.027% with A6 vs
−0.04%; 95% CI −0.041, −0.035% with B2, respectively;
The results are expressed as the changes in response to the A6 Fig. 2a and b). Similarly, BMI and waist circumference de-
and B2 regimens, presented as means with 95% CIs (Fig. 2 and creased with both regimens ( p < 0.001), more with B2
Table 2). The factors ‘period’ and ‘sequence’ (the order of the ( p<0.001; −0.82 kg/m2; 95% CI −0.94, −0.69 kg/m2 with
Diabetologia (2014) 57:1552–1560 1557

Table 2 Changes in anthropometric and laboratory variables in response to regimens of six (A6) and two (B2) meals a day

Variables Six meals a day (A6) Two meals a day (B2) p values

Dietary intake – energy (kJ/day) −1,590 (−1,970 to −1,054)*** −1,757 (−2,105 to −1,201)*** 0.731
Dietary intake – fat (g/day) −35.0 (−43.7 to −27.9)*** −38.2 (−44.4 to −32.2)*** 0.921
Dietary intake – carbohydrates (g/day) −19.7 (−32.4 to −3.1)*** −23.2 (−37.8 to −8.5)*** 0.873
Dietary intake – protein (g/day) +0.9 (−3.7 to +5.6) −4.6 (−8.7 to +1.1) 0.637
Step count (steps/month) +2,092 (+879 to +3,493)* +2,213 (+996 to +3,612)* 0.876
BMI (kg/m2) −0.82 (−0.94 to −0.69)*** −1.23 (−1.4 to −1.17)*** <0.001
Waist circumference (cm) −1.37 (−2.01 to −0.73)*** −5.14 (−5.78 to −4.50)*** <0.001
Fasting plasma glucose (mmol/l) −0.47 (−0.57 to −0.36)*** −0.78 (−0.89 to −0.68)*** 0.004
Fasting immunoreactive insulin (pmol/l) −0.69 (−1.18 to −0.21)* −0.75 (−1.23 to −0.27)* 0.910
Triacylglycerols (mmol/l) −0.28 (−0.39 to −0.17)** −0.17 (−0.28 to −0.06)* 0.300
Total cholesterol (mmol/l) −0.05 (−0.13 to +0.04) −0.07 (−0.15 to +0.01) 0.730
HDL-cholesterol (mmol/l) +0.016 (−0.006 to +0.038) +0.003 (−0.019 to +0.025) 0.570
LDL-cholesterol (mmol/l) −0.08 (−0.15 to −0.01)* −0.06 (−0.13 to −0.01) 0.823
Insulin secretion at reference level (pmol min−1 m−2) +22.9 (+11.7 to +34.0)* +20.0 (+8.7 to +31.2)* 0.795
Glucose sensitivity (pmol min−1 m−2 mmol−1 l−1) +5.8 (+2.3 to +9.5)* +5.9 (+2.3 to +9.5)* 0.991
Rate sensitivity (pmol m−2 mmol−1 l−1) −141.9 (−248.2 to −36.8)* −251.5 (−358.9 to −145.3)* 0.303
Potentiation factor (dimensionless) −0.034 (−0.059 to −0.009)* −0.038 (−0.062 to −0.013)* 0.890

Data are mean ± 95% CI

Listed p values are from 2×2 crossover ANOVA for the factor ‘treatment’
Significant changes during the regimens are indicated by: *p<0.05; **p<0.01; ***p<0.001

A6 vs −1.23 kg/m2; 95% CI −1.4, −1.17 kg/m2 with B2; and with A6 vs +0.52 ml kg − 1 min − 1 ; 95% CI +0.30,
−1.37 cm; 95% CI −2.01, −0.73 cm with A6 vs −5.14 cm; +0.74 ml kg−1 min−1 with B2; p=0.8; Fig. 2g). OGIS in-
95% CI −5.78, −4.50 cm in B2, respectively; Table 2). creased in both regimens ( p<0.01), more with B2 ( p=0.01;
+8.2 ml min−1 m−2; 95% CI +3.4, +13.1 ml min−1 m−2 with
Glycaemic control A6 vs +21 ml min−1 m−2; 95% CI +16.1, +26.0 ml min−1 m−2
with B2; Fig. 2h).
Fasting plasma glucose decreased under both regimens
( p<0.001), more with B2 ( p=0.004; −0.47 mmol/l; 95% CI
−0.57, −0.36 mmol/l with A6 vs −0.78 mmol/l; 95% CI −0.89, Beta cell function
−0.68 mmol/l with B2; Fig. 2c). Fasting C-peptide decreased
in both regimens, more with B2 ( p=0.04; −0.049 nmol/l; 95% Insulin secretion at the reference level and glucose sensitivity
CI −0.091, −0.006 nmol/l with A6 vs p<0.001; −0.14 nmol/l increased ( p<0.05) comparably with both regimens. Rate
95% CI −0.181, −0.099 nmol/l with B2; p=0.04; Fig. 2d). sensitivity and potentiation factor also decreased ( p<0.05)
Fasting immunoreactive insulin decreased ( p<0.04) compa- comparably (Table 2).
rably with both regimens (−0.69 pmol/l; 95% CI −1.18, −0.21
pmol/l with A6 vs −0.75 pmol/l; 95% CI −1.23, −0.27 pmol/l Plasma lipids
with B2; p=0.9; Table 2). Fasting glucagon decreased with B2
( p<0.001; −343 pg/ml; 95% CI −375, −311 pg/ml), whereas Triacylglycerols and LDL-cholesterol decreased comparably
it increased ( p=0.04; +53 pg/ml; 95% CI +22, +84 pg/ml) under both regimens. No significant change in total or
with A6 ( p<0.001; Fig. 2e). HbA1c decreased ( p<0.001) HDL-cholesterol was observed in either regimen (Table 2).
comparably with both regimens (−0.23%; 95% CI −0.27,
−0.19% with A6 vs −0.25%; 95% CI −0.29, −0.20% with
B2; p=0.08; Fig. 2f). REE

Whole-body insulin sensitivity REE decreased under both regimens ( p<0.001), with a trend
toward a greater decrease with A6 (−453.1 kJ/day; 95% CI
MCR increased ( p<0.001) comparably with both regimens −524.3, −382.8 kJ/day with A6 vs −379.9 kJ/day; 95% CI
(+0.45 ml kg−1 min−1; 95% CI +0.24, +0.67 ml kg−1 min−1 −449.8, −310.9 kJ/day with B2; p=0.3; Fig. 2i).
1558 Diabetologia (2014) 57:1552–1560

Correlations Studies involving individuals with type 2 diabetes are

rather limited in both length and sample size. A longer term
The decrease in HFC showed a strong positive correlation study, comparing the effect of three and nine meals daily (each
with the decrease in fasting plasma glucose (r = +0.56; period lasting 4 weeks), in 13 patients with type 2 diabetes, did
p<0.001). This association remained significant even after not confirm the beneficial effects of increased meal frequency
adjustment for changes in BMI (r=+0.28; p=0.05). [39]. It has recently been demonstrated that, for glycaemic
Changes in glucose sensitivity and OGIS correlated nega- control, eating fewer larger meals rich in fibre instead of more
tively with changes in HFC (r=−0.28; p=0.02 and r=−0.47; smaller ones may be more beneficial for type 2 diabetic
p<0.001, respectively). After adjustment for changes in BMI, patients [10].
the correlations were no longer significant. Ours is the first study into the effect of meal frequency on
insulin resistance and HFC. MCR increased in response to
both regimens, with a trend toward a greater increase with B2.
Discussion The difference between the two regimens was not significant,
probably due to a limited number of participants. OGIS in-
Principal findings creased in response to both regimens, more for B2. The
increased insulin sensitivity positively influenced the decrease
This randomised crossover 24 week study examined the effect in fasting plasma glucose and HFC (although the HbA1c level
of frequency of meals on body weight, HFC, insulin resistance decreased comparably in both regimens) or, conversely,
and beta cell function in type 2 diabetic patients. A compar- decreased HFC may have led to increased insulin sensitivity,
ison of the effect of six vs two meals (breakfast and lunch) because HFC is typically associated with insulin resistance
with the same daily energy restriction (−2,092 kJ/day) and (independent of BMI) [40], metabolic syndrome, type 2 dia-
macronutrient content, each regimen lasting 12 weeks, dem- betes and subclinical atherosclerosis [41]. In this context, the
onstrated a superior effect of B2 on body weight, HFC, fasting greater reduction in HFC in the B2 regimen is one of the most
plasma glucose, C-peptide, glucagon and OGIS. The effect of important results of our study.
meal frequency on MCR and beta cell function was not We are also the first to observe an effect on fasting plasma
significant. concentrations of glucagon with a B2 regimen. Inappropriate-
ly elevated plasma glucagon concentrations play a role in
Findings in relation to other research dysregulated hepatic glucose production and abnormal glu-
cose homeostasis in type 2 diabetes [42]. There is no treatment
The superior effect of B2 on most of the variables studied that specifically decreases glucagon levels. However, incretin-
supports our hypothesis and previous lines of evidence from based agents reduce plasma glucagon, which contributes to
animal models, observational studies and randomised trials their action to lower blood glucose [43]. In this regard, a
[5, 7, 8, 12, 14]. decrease of glucagon with B2 in our study is a very positive
Our results are in strong agreement with animal studies, finding.
which have demonstrated the glucose-lowering effects of
intermittent fasting regimens (such as every other day fasting Possible mechanisms
or fasting for 2 days a week). These regimens reduce blood
glucose and insulin concentrations, improve glucose tolerance The mechanisms responsible for a greater weight loss with the
[4, 28], mainly due to increased insulin sensitivity [29], and B2 regimen may be the trend toward a smaller decrease in
extend the lifespan of laboratory animals [30]. REE with B2 that we measured (although the difference
Our data contradict the widely held opinion that eating between regimens did not reach the threshold for statistical
more frequently is healthier than eating less frequent larger significance), together with a greater thermogenic response of
meals. Some studies have suggested that people who consume larger meals, as documented by others [11].
more snacks are less likely to be obese [31], but other large Furthermore, some of the positive effects of intermittent
prospective studies have demonstrated that frequent snacking fasting on glucose homeostasis may be mediated by the ner-
may lead to weight gain [32] and an increased risk of type 2 vous system, mainly by the increased production of brain-
diabetes [33, 34] because of the higher energy intake, mainly derived neurotrophic factor, which increases the resistance of
from added sugars [35]. Furthermore, the reported benefits of neurones to dysfunction and degeneration in animal models
more frequent meals are usually associated with meal frequen- [3]. It is possible that overnight fasting between lunch and
cies exceeding those which might be translated into practical breakfast the next day elicited some of these effects.
recommendations [36]. In fact, some of the early experiments Periods of fasting between meals may be even more im-
with frequency of eating [37, 38] only underline the impor- portant than the composition of the diet. In one experimental
tance of fibre and low glycaemic index foods in the diet. study, the timing of meals led to increased insulin sensitivity
Diabetologia (2014) 57:1552–1560 1559

and decreased body weight in spite of the high fat content of macronutrient content but also the frequency and timing of
the consumed diet [6]. The distribution of the meals is another food. Further larger scale, long-term studies are essential
important factor. Eating meals later in the day may also before offering recommendations in terms of meal frequency.
adversely influence the success of a weight loss therapy. This
difference in weight loss success was not explained by differ- Acknowledgements We thank the 54 participants for their cooperation,
ences in caloric intake, macronutrient distribution or energy as well as our staff from the Diabetes Centre, Institute for Clinical and
expenditure [12]. A potential mechanism explaining this dif- Experimental Medicine, Prague, Czech Republic: the registered dietitians
(V. Havlova and R. Milatova) for providing both group and individual
ference is that the timing of food intake can influence the nutrition counselling and study nurses (D. Lapesova, D. Sisakova,
circadian system [44]. The circadian system must continuous- D. Kobrova, J. Purrova and B. Vodickova) for conducting the procedures.
ly adapt to and synchronise our physiology with the environ- We thank J. Sievenpiper from the Clinical Nutrition and Risk Factor
ment [45]. A genetic variance in clock genes may be important Modification Centre, St Michael's Hospital, Toronto, Canada for great
help with text corrections. Some of the data were presented as an abstract
in meal timing, possibly in part by changes in the recently at the 72nd and 73rd Scientific Sessions of the ADA and 48th and 49th
demonstrated circadian control of hunger and appetite [46]. EASD Annual Meetings in 2012 and 2013.
Another recent study demonstrated that a high carbohy-
drate and protein breakfast may prevent weight regain by Funding This study was supported by grant NT/11238-4 from Ministry
reducing diet-induced compensatory changes in hunger, crav- of Health, Prague, Czech Republic and by the Grant Agency of Charles
ings and ghrelin suppression [47]. Although the mechanisms University - GAUK No 702312.
linking the timing of meals and the regulation of body weight
are unknown, satiety hormones, such as leptin or ghrelin, may Duality of interest The authors declare that there is no duality of
be involved [48]. Changes in the levels of these hormones by interest associated with this manuscript.
circadian misalignment could influence energy intake and
Contribution statement HK, LB and TP designed the study, wrote the
expenditure [49]. grant application, recruited the patients, collected the data and wrote the
manuscript. HM, OO, JT, VS, LK, MD, MHa and AT were involved in
Weaknesses the acquisition and analysis of the data. MHi carried out the statistical
analyses and interpretation of data. All authors had full access to data
and revised and approved the manuscript for publication. The
The short duration of our study and provision of food pre- guarantor is TP.
cludes a generalisation of our study to free-living conditions.
We provided all the meals during the whole study for one half Open Access This article is distributed under the terms of the Creative
of the participants in order to ensure the best possible compli- Commons Attribution License which permits any use, distribution, and
ance, yet we have to admit the possibility of a reduced energy reproduction in any medium, provided the original author(s) and the
source are credited.
intake with the B2 regimen, even though the energy intake
reported by our participants was similar in both regimens. The
dropout rates were also comparable for both regimens.
The sample of diabetic patients in our study was not rep- References
resentative: the duration of diabetes in our patients was quite
short; all the participants were being treated by oral 1. Visscher TL, Seidell JC (2001) The public health impact of obesity.
hypoglycaemic agents, with motivation not to initiate insulin Annu Rev Public Health 22:355–375
therapy, and they were willing to make substantial changes in 2. Mattson MP (2005) The need for controlled studies of the effects of
meal frequency on health. Lancet 365:1978–1980
their lifestyle. 3. Mattson MP (2005) ENERGY intake, meal frequency, and health: a
neurobiological perspective. Annu Rev Nutr 25:237–260
Conclusions and research needs 4. Anson RM, Guo Z, de Cabo R et al (2003) Intermittent fasting
dissociates beneficial effects of dietary restriction on glucose
metabolism and neuronal resistance to injury from calorie intake.
In conclusion, a hypoenergetic diet (a restriction of 2,092 kJ/ Proc Natl Acad Sci U S A 100:6216–6220
day) consumed as breakfast and lunch reduced body weight, 5. Hatori M, Vollmers C, Zarrinpar A et al (2012) Time-restricted
HFC, fasting plasma glucose, C-peptide and glucagon, and feeding without reducing caloric intake prevents metabolic diseases
increased calculated insulin sensitivity (OGIS), more than the in mice fed a high-fat diet. Cell Metab 15:848–860
6. Sherman H, Genzer Y, Cohen R et al (2012) Timed high-fat diet
same hypoenergetic diet (a similar restriction of 2,092 kJ/day) resets circadian metabolism and prevents obesity. FASEB J 26:3493–
divided into six, more frequent, meals. MCR and beta cell 3502
function improved comparably in both regimens. These re- 7. Howarth NC, Huang TT-K, Roberts SB et al (2007) Eating patterns
sults suggest that eating two larger meals a day (breakfast and and dietary composition in relation to BMI in younger and older
adults. Int J Obes 2005 31:675–684
lunch) may be more beneficial for patients with type 2 diabe- 8. Duval K, Strychar I, Cyr M-J et al (2008) Physical activity is a
tes than six smaller meals during the day. Novel therapeutic confounding factor of the relation between eating frequency and
strategies should incorporate not only the energy and body composition. Am J Clin Nutr 88:1200–1205
1560 Diabetologia (2014) 57:1552–1560

9. Bachman JL, Raynor HA (2012) Effects of manipulating eating 30. Goodrick CL, Ingram DK, Reynolds MA et al (1982) Effects of
frequency during a behavioral weight loss intervention: a pilot ran- intermittent feeding upon growth and life span in rats. Gerontology
domized controlled trial. Obesity (Silver Spring) 20:985–992 28:233–241
10. Fernemark H, Jaredsson C, Bunjaku B et al (2013) A randomized 31. Keast DR, Nicklas TA, O’Neil CE (2010) Snacking is associated with
cross-over trial of the postprandial effects of three different diets in reduced risk of overweight and reduced abdominal obesity in ado-
patients with type 2 diabetes. PLoS One 8:e79324 lescents: National Health and Nutrition Examination Survey
11. Tai MM, Castillo P, Pi-Sunyer FX (1991) Meal size and frequency: (NHANES) 1999–2004. Am J Clin Nutr 92:428–435
effect on the thermic effect of food. Am J Clin Nutr 54:783–787 32. van der Heijden AAWA, Hu FB, Rimm EB, van Dam RM (2007) A
12. Garaulet M, Gómez-Abellán P, Alburquerque-Béjar JJ et al (2013) Timing prospective study of breakfast consumption and weight gain among
of food intake predicts weight loss effectiveness. Int J Obes (Lond) U.S. men. Obesity (Silver Spring) 15:2463–2469
37:604–611 33. Mekary RA, Giovannucci E, Willett WC et al (2012) Eating patterns
13. Ruge T, Hodson L, Cheeseman J et al (2009) Fasted to fed trafficking and type 2 diabetes risk in men: breakfast omission, eating frequency,
of fatty acids in human adipose tissue reveals a novel regulatory step and snacking. Am J Clin Nutr 95:1182–1189
for enhanced fat storage. J Clin Endocrinol Metab 94:1781–1788 34. Mekary RA, Giovannucci E, Cahill L et al (2013) Eating patterns and
14. Purslow LR, Sandhu MS, Forouhi N et al (2008) Energy intake at type 2 diabetes risk in older women: breakfast consumption and
breakfast and weight change: prospective study of 6,764 middle-aged eating frequency. Am J Clin Nutr 98:436–443
men and women. Am J Epidemiol 167:188–192 35. Larson N, Story M (2013) A review of snacking patterns among
15. Mann JI, de Leeuw I, Hermansen K et al (2004) Evidence-based children and adolescents: what are the implications of snacking for
nutritional approaches to the treatment and prevention of diabetes weight status? Child Obes Print 9:104–115
mellitus. Nutr Metab Cardiovasc Dis 14:373–394 36. Mann J (1997) Meal frequency and plasma lipids and lipoproteins.
16. Ferrannini E (1988) The theoretical bases of indirect calorimetry: a Br J Nutr 77(Suppl 1):S83–S90
review. Metabolism 37:287–301 37. Jenkins DJ, Wolever TM, Vuksan V et al (1989) Nibbling
17. Hagströmer M, Oja P, Sjöström M (2006) The International Physical versus gorging: metabolic advantages of increased meal frequency.
Activity Questionnaire (IPAQ): a study of concurrent and construct N Engl J Med 321:929–934
validity. Public Health Nutr 9:755–762 38. Fabry P, Hejl Z, Fodor J et al (1964) The frequency of meals. Its
18. Baecke JA, Burema J, Frijters JE (1982) A short questionnaire for the relation to overweight, hypercholesterolaemia, and decreased
measurement of habitual physical activity in epidemiological studies. glucose-tolerance. Lancet 2:614–615
Am J Clin Nutr 36:936–942 39. Arnold L, Mann JI, Ball MJ (1997) Metabolic effects of alterations in
19. Pelikánová T, Smrcková I, Krízová J et al (1996) Effects of insulin meal frequency in type 2 diabetes. Diabetes Care 20:1651–1654
and lipid emulsion on renal haemodynamics and renal sodium han- 40. Seppälä-Lindroos A, Vehkavaara S, Häkkinen A-M et al (2002) Fat
dling in IDDM patients. Diabetologia 39:1074–1082 accumulation in the liver is associated with defects in insulin
20. Hájek M, Dezortová M, Wagnerová D et al (2011) MR spectroscopy suppression of glucose production and serum free fatty acids inde-
as a tool for in vivo determination of steatosis in liver transplant pendent of obesity in normal men. J Clin Endocrinol Metab 87:3023–
recipients. Magma N Y N 24:297–304 3028
21. Longo R, Pollesello P, Ricci C et al (1995) Proton MR spectroscopy 41. Juárez-Rojas JG, Medina-Urrutia AX, Jorge-Galarza E et al (2013)
in quantitative in vivo determination of fat content in human liver Fatty liver increases the association of metabolic syndrome with
steatosis. J Magn Reson Imaging 5:281–285 diabetes and atherosclerosis. Diabetes Care 36:1726–1728
22. van Cauter E, Mestrez F, Sturis J, Polonsky KS (1992) Estimation of 42. D’Alessio D (2011) The role of dysregulated glucagon secretion in
insulin secretion rates from C-peptide levels. Comparison of individ- type 2 diabetes. Diabetes Obes Metab 13(Suppl 1):126–132
ual and standard kinetic parameters for C-peptide clearance. Diabetes 43. Edwards KL, Stapleton M, Weis J, Irons BK (2012) An update in
41:368–377 incretin-based therapy: a focus on glucagon-like peptide-1 receptor
23. Mari A, Tura A, Gastaldelli A, Ferrannini E (2002) Assessing insulin agonists. Diabetes Technol Ther 14:951–967
secretion by modeling in multiple-meal tests: role of potentiation. 44. Garaulet M, Madrid JA (2010) Chronobiological aspects of nutrition,
Diabetes 51(Suppl 1):S221–S226 metabolic syndrome and obesity. Adv Drug Deliv Rev 62:967–978
24. Mari A, Schmitz O, Gastaldelli A et al (2002) Meal and oral glucose 45. Bienertová-Vasků J, Bienert P, Forejt M et al (2010) Genotype x
tests for assessment of beta -cell function: modeling analysis in nutrient association of common polymorphisms in obesity-related
normal subjects. Am J Physiol Endocrinol Metab 283:E1159–E1166 genes with food preferences and time structure of energy intake.
25. Kahleova H, Mari A, Nofrate V et al (2012) Improvement in β-cell Br J Nutr 103:352–359
function after diet-induced weight loss is associated with 46. Scheer FAJL, Morris CJ, Shea SA (2013) The internal circadian
decrease in pancreatic polypeptide in subjects with type 2 diabetes. clock increases hunger and appetite in the evening independent of
J Diabetes Complicat 26:442–449 food intake and other behaviors. Obesity (Silver Spring) 21:421–
26. Meloun M, Hill M, Militký J, Kupka K (2000) Transformation in the 423
PC-aided biochemical data analysis. Clin Chem Lab Med 38:553–559 47. Jakubowicz D, Froy O, Wainstein J, Boaz M (2012) Meal timing and
27. Meloun M, Militký J, Hill M, Brereton RG (2002) Crucial problems composition influence ghrelin levels, appetite scores and weight loss
in regression modelling and their solutions. Analyst 127:433–450 maintenance in overweight and obese adults. Steroids 77:323–331
28. Pedersen CR, Hagemann I, Bock T, Buschard K (1999) Intermittent 48. Arble DM, Bass J, Laposky AD et al (2009) Circadian timing of food
feeding and fasting reduces diabetes incidence in BB rats. intake contributes to weight gain. Obesity (Silver Spring) 17:2100–
Autoimmunity 30:243–250 2102
29. Anson RM, Jones B, de Cabod R (2005) The diet restriction para- 49. Ma Y, Bertone ER, Stanek EJ 3rd et al (2003) Association between
digm: a brief review of the effects of every-other-day feeding. eating patterns and obesity in a free-living US adult population.
Age (Dordr) 27:17–25 Am J Epidemiol 158:85–92