Effect of intravenous glucose and euglycemic insulin
infusions on short-term appetite and food intake
IAN M. CHAPMAN,1 ELIZABETH A. GOBLE,1 GARY A. WITTERT,1
JOHN E. MORLEY,2 AND MICHAEL HOROWITZ1
1Department of Medicine, Royal Adelaide Hospital, Adelaide, South Australia, Australia 5000; and
2Department of Geriatrics, St. Louis University and Geriatric Research Education and Clinical
Center, St. Louis Veterans Affairs Medical Center, St. Louis, Missouri 63104
glucoprivation; feeding; hunger
in the regulation of feeding in
humans is controversial. Although exogenous insulin
has been reported to increase food intake in both
animals and humans (3, 15, 26, 32), these observations
may be attributable to glucoprivation and/or a reduction in blood glucose concentrations (relative or absolute hypoglycemia), rather than the effects of insulin
itself. In particular, 2-deoxy-D-glucose, which inhibits
glucose utilization and induces intracellular glucopenia
without affecting plasma insulin concentrations, also
increases food intake in humans (35). There is considerable evidence, derived to our knowledge entirely from
animal studies, that insulin may inhibit food intake.
Central and peripheral administration of insulin reduces food intake in both rodents (4) and baboons (39,
40), provided hypoglycemia is prevented. The physiological significance of these observations is uncertain, as
the majority of these studies have involved either the
THE ROLE OF INSULIN
R596
direct injection of insulin into the brain or peripheral
administration in pharmacological doses for prolonged
periods of time. The observation in rats that central
administration of insulin antibodies is associated with
an increase in food intake (33) is more persuasive
evidence in support of a central appetite-suppressive
effect of insulin in that species.
We are not aware of any studies that demonstrate a
suppressive effect of insulin on feeding in humans.
Rodin et al. (29) performed intravenous hyperinsulinemic hypo- and hyperglycemic clamp studies, the latter
using a glucose infusion to stimulate endogenous insulin release, and found that both treatments stimulated
short-term appetite and food intake in healthy young
adults. They interpreted their results to indicate that
insulin stimulates human feeding. Holt and Miller (18)
manipulated plasma insulin concentrations physiologically, and higher levels were associated with lessened
satiety. Woo et al. (38), on the other hand, found that
appetite and food intake in healthy young adults were
not affected by intravenous nonhypoglycemic insulin
infusions.
Insulin secretion is stimulated more by enteral than
intravenous glucose administration due to the incretin
effect (22, 25). Enteral glucose administration, whether
by the oral or small intestinal route, also suppresses
appetite and subsequent food intake more than intravenous glucose administration (28), even when the resultant increase in blood glucose concentration is similar
(22). This enhanced appetite suppression could potentially be mediated by the release of incretins, such as
glucagon-like peptide-I (GLP-I) and gastric inhibitory
polypeptide, or by other gastrointestinal hormones (8,
12, 20, 36). However, in view of the animal evidence
implicating insulin as a satiety agent, it could also be
due to the greater insulin release associated with
enteral glucose administration. The present study was
performed to investigate the short-term effects of insulin on human feeding behavior at concentrations comparable to those seen after a meal.
METHODS
Subjects
Fourteen healthy nonsmoking young adult subjects (12
men and 2 woman), 20–33 yr old, with body mass indexes of
20.8–26.7 kg/m2 (23.6 6 1.9, mean 6 SD) were studied. All
subjects had a score of 10 or less (mean, 4.9 6 2.6) on the
eating restraint factor of the Eating Inventory Questionnaire
(34), indicating that they were not restrained eaters, and
none was dieting. Both female subjects were taking a combined oral contraceptive pill; otherwise, subjects were not
0363-6119/98 $5.00 Copyright r 1998 the American Physiological Society
Downloaded from http://ajpregu.physiology.org/ by 10.220.32.247 on June 18, 2017
Chapman, Ian M., Elizabeth A. Goble, Gary A. Wittert, John E. Morley, and Michael Horowitz. Effect of
intravenous glucose and euglycemic insulin infusions on
short-term appetite and food intake. Am. J. Physiol. 274
(Regulatory Integrative Comp. Physiol. 43): R596–R603,
1998.—To investigate the short-term effects of insulin on
feeding, 14 fasting, young adults received 150-min euglycemic intravenous infusions of control (C), low-dose (LD, 0.8
mU · kg21 · min21 ), and high-dose (HD, 1.6 mU · kg21 · min21 )
insulin and ate freely from a buffet meal during the last 30
min. Steady-state preprandial plasma insulin concentrations
were 5.9 6 0.7 (C), 47 6 2 (LD), and 95 6 6 (HD) µU/ml and
increased 56–59 µU/ml during the meal. No effect of treatment type on hunger or fullness ratings, duration of eating, or
the weight, energy content (1,053 6 95 kcal, C; 1,045 6 101
kcal, LD; 1,066 6 107 kcal, HD; P 5 0.9), and composition of
food eaten was observed. On a fourth study day, 12 of the
subjects received an intravenous infusion of glucose only (Glc)
that was identical to the glucose infusion on their HD insulin
day. Mean venous glucose concentration was 9.3 6 0.5 mmol
[P , 0.001 vs. C (5.3 6 0.1), LD (5.2 6 0.2), and HD (5.2 6
0.2)], and plasma insulin increased to 45 6 2.3 µU/ml at the
start and 242 6 36 µU/ml at the end of the meal. Energy
intake during the meal was (,15%) reduced (1,072 6 97 kcal,
C; 1,086 6 102 kcal, LD; 1,088 6 105 kcal, HD; 919 6 115
kcal, Glc; P , 0.05 Glc vs. C, LD, and HD). Plasma insulin
normally increases to ,100 µU/ml after a mixed meal in lean
subjects. Therefore, in the absence of altered blood glucose
concentrations, physiological concentrations of insulin are
unlikely to play a role in meal termination and the short-term
control of appetite.
EFFECT OF INSULIN ON HUMAN APPETITE AND FOOD INTAKE
taking any medications. The protocol was approved by the
Human Ethics Committee of the Royal Adelaide Hospital,
and subjects gave written informed consent.
Protocol
Fig. 1. Study design. See text for detailed explanation.
portable blood glucose meter (MediSense Companion 2 Blood
Glucose System; Medisense, Waltham, MA).
At 140 min, subjects were presented with a cold buffet meal
prepared in excess of what they would normally be expected
to eat and were instructed to eat as much as they wished (22).
The composition of the meal was explained to the subjects at
enrollment, and substitute items were arranged when requested. Each subject received the same meal on all three
study days. For 11 subjects, the meal consisted of sliced bread
(white and whole meal), nondairy spread, mayonnaise, sliced
ham, chicken, cheese, tomato and cucumber, lettuce, milk,
orange juice, strawberry yogurt, chocolate custard, vanilla ice
cream, an apple, pear, and banana. For the other three
subjects the meal was the same except that hard-boiled eggs
were offered instead of sliced ham and chicken. Subjects were
allowed 30 min to eat, during which time the glucose infusion
rate was kept constant at the 140-min value. The rate and
duration of ingestion and the total amount of food eaten were
recorded. At the end of the 30-min meal period, a further
venous blood sample was taken, a VAS was administered, and
the intravenous glucose, saline, and treatment infusions were
stopped.
Study 2. On completion of study 1, the 14 subjects were
asked if they would return for a fourth study day. Twelve
agreed and returned on the next suitable day. The subjects
were not told what treatment they would receive but that the
protocol would be similar to one of the previous three visits.
The study protocol for this ‘‘glucose-only’’ study day was
exactly the same as that of the high-dose insulin study day
except that no insulin was administered; for each subject, the
glucose infusion exactly reproduced that of their previous
high-dose insulin study day.
Assessment of appetite and food intake. Appetite was
assessed using 10-cm VAS (31) on which hunger and fullness
were quantified. Subjects were familiarized with these scales
at the beginning of the study and were instructed to make a
single vertical mark on the scale to indicate their current
feelings. The 0- and 20-min values were averaged to produce
a baseline value, and the changes in ratings from baseline
were quantified (22).
Food intake was calculated by weighing each food item
before and after the meal. The DIET/1 Nutrient Calculation
software package (Xyris Software; Highgate Hill, Queensland, Australia) was used to determine the energy intake
(kcal) and macronutrient composition (%protein, %carbohydrate, and %fat) of the meal.
Assays. Plasma insulin was measured using the Abbott
Imx Microparticle Enzyme Immunoassay (Abbott Laboratories, Diagnostic Division, Dainabot, Tokyo, Japan). The sensitivity of the assay (concentration at 2 SD from the zero
standard) was 1.0 µU/ml. Samples with a value above that of
the highest standard (300 µU/ml) were reassayed at a dilution of 1:2. The intra-assay coefficients of the assay were 4%
at 8.3 µU/ml, 2.9% at 40.4 µU/ml, and 2.5% at 121.7 µU/ml.
The interassay coefficients of variation were 3.7% at 8 µU/ml,
2.8% at 40 µU/ml, and 3.7% at 120 µU/ml.
Statistical analysis. Separate analyses were performed on
studies 1 and 2. Twelve of the fourteen subjects took part in
both studies and three of their four study days were analyzed
as part of both. Differences in appetite ratings at baseline and
parameters of food intake were assessed using repeatedmeasures one-way analysis of variance (ANOVA). Appetite
ratings, blood glucose, and plasma insulin concentrations
during the studies were assessed using repeated-measures
two-way ANOVA, with treatment and time as the factors.
Multiple-comparison tests were performed using the StudentNewman-Keuls Test. All calculations were performed using
Downloaded from http://ajpregu.physiology.org/ by 10.220.32.247 on June 18, 2017
Study 1. Subjects were studied on the following three
occasions, at least 10 days apart: 1) control, 2) 0.8
mU · kg21 ·min21 insulin (low dose), and 3) 1.6 mU·kg21 ·min21
insulin (high dose). Studies were performed in random order
and single-blind fashion. For females, each study was performed between days 1 and 7 of the oral contraceptive pill
treatment cycle. Subjects were advised not to modify their
diet throughout the study period and to avoid alcohol and
strenuous exercise for at least 24 h before each study.
Subjects fasted overnight (nothing except water after 2100 h)
and attended the research center the next morning at 0800 h.
A cannula was inserted into an arm vein to administer
infusions. A second cannula was inserted into a vein of the
opposite arm for collection of blood samples; this arm was
kept warm using a heating pad to ‘‘arterialize’’ the blood.
The experimental protocol is summarized in Fig. 1. After
placement of the cannulas, subjects rested quietly for 30 min.
They were then (0 min) asked to record their feelings of
hunger and fullness on visual analogue scales (VAS). VAS
were administered every 20 min thereafter until 140 min,
when a meal was given. Further VAS were administered at
the end of the meal (170 min) and at the end of the study (200
min). Venous blood samples were obtained at the same time
points for subsequent measurement of plasma insulin and at
additional time points for measurement of blood glucose (see
below).
The intravenous infusion regimens were as follows. From
15 to 170 min, 0.9% saline containing 25 mmol potassium
chloride/l was infused intravenously at 200 ml/h. For 2.5 h,
from 20 to 170 min, insulin in Haemaccel (Behringwerke,
Marburg, Germany) or Haemaccel alone (control) was infused
intravenously. The infusion rate was 0.9 ml · kg body wt21 · h21
for the first 5 min, 0.7 ml · kg21 · h21 for the next 5 min, and
then 0.5 ml · kg21 · h21 until 170 min. On the low- and highdose insulin infusion days, regular insulin (Novo-Nordisk
Pharmaceutical, North Rocks, New South Wales, Australia)
was added to the Haemaccel to produce steady-state insulin
infusion rates (30–170 min) of 0.8 and 1.6 mU · kg21 · min21,
respectively. Twenty-five percent glucose was infused intravenously from 25 to 170 min. The glucose infusion rate was
varied according to the results of bedside blood glucose
measurements to maintain glucose levels between baseline
and baseline plus 10%. The baseline glucose concentration
was defined as the mean of the 10- and 15-min values.
Bedside testing of blood glucose was performed every 5 to 10
min from 10 to 140 min, at 170 min, and at 200 min, using a
R597
R598
EFFECT OF INSULIN ON HUMAN APPETITE AND FOOD INTAKE
the SigmaStat program (Jandel Scientific Software, San
Rafael, CA). Results are reported as means 6 SE. A P value
of , 0.05 was considered significant in all analyses.
RESULTS
The study protocol was well tolerated in all subjects,
and there were no adverse events.
Study 1
Fig. 2. Mean 6 SE venous glucose concentrations (top) and plasma
insulin concentrations (bottom) during euglycemic iv infusions of
control (j), low-dose insulin (0.8 mU · kg21 · min21; k), and high-dose
insulin (1.6 mU · kg21 · min21; l); n 5 14 subjects.
Fig. 3. Change from baseline in ratings of hunger (top) and fullness
(bottom) during euglycemic iv infusions of control (j), low-dose
insulin (k), and high-dose insulin (l); n 5 14. Significant effect of
time on hunger and fullness [P , 0.01 by analysis of variance
(ANOVA)] but no significant effect of treatment on either hunger or
fullness.
ies (59 6 9 control, 56 6 12 low dose, 57 6 9 µU/ml high
dose, P . 0.05).
Baseline (pretreatment) ratings of hunger (F 5 1.8,
P 5 0.2) and fullness (F 5 0.84, P 5 0.45) did not differ
significantly between the three treatment days. Changes
in hunger ratings during the three treatment infusions
(baseline to 170 min) are shown in Fig. 3, top. There
was an effect of time on hunger (F 5 47.2, P , 0.0001);
subjects rated themselves as more hungry immediately
before the meal and less hungry after the meal than at
baseline. There was no effect of treatment type on
hunger rating (F 5 0.5, P 5 0.6) and no interaction
between treatment type and time (F 5 0.8, P 5 0.7).
When the analysis was restricted to the time period
between baseline and the start of the meal (140 min),
there was also a significant effect of time (F 5 4.4, P 5
0.008), but not treatment type (F 5 0.26, P 5 0.77), and
no interaction between time and treatment type (F 5
0.72, P 5 0.73).
Changes in ratings of fullness during the treatment
infusions are shown in Fig. 3, bottom. There was an
effect of time (F 5 95, P , 0.0001), with subjects rating
themselves as being less full immediately before and
more full after the meal than at baseline. There was no
effect of treatment type on fullness rating (F 5 1.7, P 5
0.2) and no interaction between treatment type and
time (F 5 1.1, P 5 0.33). When only the period from
baseline to the start of the meal was analyzed, similar
effects were observed (time, F 5 4, P 5 0.001; treat-
Downloaded from http://ajpregu.physiology.org/ by 10.220.32.247 on June 18, 2017
The steady-state glucose infusion rates (mean of 110to 170-min values) were 0.7 6 0.1, 7 6 0.7, and 9 6 0.7
mg · kg21 · min21 on the control, low-dose, and high-dose
insulin infusion days, respectively. Before the meal,
blood glucose concentrations did not differ significantly
between the three treatments; all mean concentrations
were between 4.8 6 0.2 and 5.4 6 0.2 mmol/l (Fig. 2,
top). At the end of the meal period (170 min), glucose
concentrations were slightly lower on both insulin
infusion days than on the control day (control 6.4 6 0.3,
low dose 5.8 6 0.25, high dose 5.6 6 0.3 mmol/l, P 5
0.02 control vs. high dose and control vs. low dose, P .
0.05 high dose vs. low dose).
On the control day, plasma insulin concentrations
were ,10 µU/ml until the start of the meal and then
increased with food ingestion to 65 6 9 at the end of the
meal period and 81 6 8 µU/ml 30 min later (Fig. 2,
bottom). Plasma insulin concentrations increased within
20 min of the commencement of the low- and high-dose
insulin infusions to premeal steady-state levels of 47 6
2 and 95 6 6 µU/ml, respectively (mean of 40- to
140-min values). The increases in plasma insulin concentrations during the meal period (140 to 170 min)
were not significantly different between the three stud-
EFFECT OF INSULIN ON HUMAN APPETITE AND FOOD INTAKE
R599
ment, F 5 1.8, P 5 0.18; time 3 treatment, F 5 1.2, P 5
0.3).
No subject ate all of the food offered, and none ate for
the full 30 min allotted. The time taken to eat the meal
and the weight, energy content, and macronutrient
composition of food eaten did not differ significantly
between any of the three treatments (Table 1). There
was ,2% difference between the highest and lowest
mean energy intakes in response to these three treatments.
Study 2
Fig. 4. Mean 6 SE venous glucose concentrations (top) and plasma
insulin concentrations (bottom) during iv infusion of 25% glucose (s)
and euglycemic iv infusions of control (j), low-dose insulin (k), and
high-dose insulin (l); n 5 12.
also a significant effect of time (F 5 4.3, P 5 0.001), but
not of treatment type (F 5 0.34, P 5 0.87), and no
interaction between time and treatment type (F 5 0.6,
P 5 0.9).
Table 1. Details of test meal consumption in 14 young
adult subjects during control conditions
and euglycemic iv infusion of insulin
Insulin, mU · kg21 · min21
Food Intake
Control
0.8
1.6
P
Duration, min
Weight, g
Energy, kcal
Carbohydrate, %
Fat, %
Protein, %
21.2 6 1.5
999 6 97
1,053 6 95
47 6 1.1
33 6 1
20 6 0.7
19.7 6 1.6
981 6 96
1,045 6 101
49 6 1.5
31 6 1.3
20 6 0.6
19.5 6 1.6
1,001 6 99
1,066 6 107
46 6 0.9
33 6 1
20 6 0.6
0.2
0.9
0.9
0.08
0.08
0.7
Values are means 6 SE. P values for comparison between groups
by repeated-measures analysis of variance (ANOVA).
Fig. 5. Change from baseline in ratings of hunger (top) and fullness
(bottom) during iv infusion of 25% glucose (s) and euglycemic iv
infusions of control (j), low-dose insulin (k), and high-dose insulin
(l); n 5 12. Significant effect of time on hunger and fullness (P , 0.01
by ANOVA) but no significant effect of treatment on either hunger or
fullness.
Downloaded from http://ajpregu.physiology.org/ by 10.220.32.247 on June 18, 2017
On the glucose-only and high-dose insulin treatment
days, 328 6 25 kcal of glucose were infused. Blood
glucose concentrations increased during glucose-only
infusion to 11.7 6 0.7 mmol/l at 100 min and declined
thereafter but, at the start of the meal, remained
approximately two times as high as on the other 3 days
(Fig. 4, top). The mean glucose concentrations during
the four treatment infusions were 5.3 6 0.1, 5.2 6 0.2,
5.2 6 0.2, and 9.3 6 0.5 mmol/l (P , 0.001, glucose only
vs. the other 3 days).
Plasma insulin concentrations increased progressively during the glucose-only infusion to 47 6 6 µU/ml
at the beginning of the meal (Fig. 4, bottom), essentially
identical to those at the same time on the low-dose
insulin infusion day (45 6 2.3 mU/ml, P . 0.05), and to
242 6 36 µU/ml at the end of the meal. The mealinduced increase in plasma insulin was more than
three times greater on the glucose-only infusion day
than the three other study days (196 6 35 vs. 59 6 11,
63 6 13, and 54 6 10 µU/ml, P , 0.0001, glucose-only
day vs. all 3 other days).
Baseline (pretreatment) ratings of hunger (P 5 0.65)
and fullness (P 5 0.87) did not differ significantly
between the four study days. Changes in hunger and
fullness ratings during the four treatment infusions
(baseline to 170 min) are shown in Fig. 5. There was an
effect of time on hunger (F 5 54.6, P , 0.0001), with
subjects rating themselves as being more hungry immediately before the meal and less hungry after the meal
than at baseline. There was no effect of treatment type
on hunger rating (F 5 0.4, P 5 0.75) and no interaction
between treatment type and time (F 5 0.6, P 5 0.9).
When the analysis was restricted to the time period
between baseline and the start of the meal, there was
R600
EFFECT OF INSULIN ON HUMAN APPETITE AND FOOD INTAKE
Relationship Between Food Intake and
Appetite Ratings
Combined analysis of all study days demonstrated an
inverse relationship between premeal (mean of 120and 140-min values) ratings of hunger and fullness (r 5
20.78, P , 0.001). There was a positive correlation
between premeal ratings of hunger and the energy
content of the food eaten during that meal (r 5 0.37,
P 5 0.006) and a negative correlation between premeal
ratings of fullness and energy intake (r 5 20.35, P 5
0.009). There was no significant correlation between
Table 2. Details of test meal consumption in 12 healthy
young adult subjects during control conditions, iv
infusion of glucose, or euglycemic iv infusion of insulin
Insulin,
mU · kg21 · min21
Food
Intake
Control
0.8
1.6
Duration,
min
21.2 6 1.5 20.1 6 1.6 19.6 6 1.3
Weight, g
997 6 103 1,017 6 97 1,026 6 96
Energy,
kcal
1,072 6 97 1,086 6 102 1,088 6 105
Carbohydrate,
%
47 6 1.3
49 6 1.7
46 6 1
Fat, %
33 6 1.1
32 6 1.4
34 6 1.1
Protein, %
20 6 0.7
19 6 0.6
20 6 0.7
Glucose
P
19.1 6 1.7 0.3
894 6 112 0.3
919 6 115 0.015*
49 6 2.1 0.5
31.5 6 1.8 0.6
19.5 6 0.8 0.5
Values are means 6 SE. Glucose infusion was same 25% glucose
infusion as on insulin (1.6 mU · kg21 · min21 ) treatment day, but no
insulin was infused. P values for comparison between groups by
repeated-measures one-way ANOVA. * P , 0.05, 25% glucose day vs.
control and 0.8 and 1.6 mU · kg21 · min21 insulin by Student-NewmanKeuls test.
the change in energy intake from that on the control
day [energy intake (treatment day) 2 energy intake
(control day; kcal)] and the change in premeal fullness
(P 5 0.47) or hunger (P 5 0.55) ratings from those on
the control day, when all three interventions (low-dose
insulin, high-dose insulin, and glucose only) were analyzed together. However, when the glucose-only day
alone was compared with the control day, there was a
significant positive correlation between the change in
energy intake and hunger ratings (r 5 0.67, P 5 0.02)
and a negative correlation between changes in energy
intake and fullness ratings, which did not quite achieve
significance (r 5 0.56, P 5 0.057).
DISCUSSION
The results of this study indicate that, under euglycemic conditions, intravenous insulin administration has
no short-term effects on appetite ratings or food intake
in young adults of normal body weight. In fasting, lean,
subjects, plasma insulin concentrations increase to a
mean peak of ,100 µU/ml, ,30–60 min after the start
of a mixed meal and then return to baseline within 2–4
h (14). In the present study, insulin infusions were
started 2 h before the test meal and were continued
throughout the meal; mean steady-state preprandial
plasma insulin concentrations on the low- and highdose insulin infusion days were 47 and 95 µU/ml,
respectively, compared with a mean peak insulin concentration of 81 µU/ml after a moderately large meal on
the control day. Thus the infusion-induced increases in
plasma insulin concentrations quite closely reproduced, in both duration and magnitude, the normal
postprandial rise. In contrast to the lack of effect on
feeding of euglycemic insulin infusions, infusion of
glucose only significantly suppressed food intake and
was associated with a nonsignificant suppression in
appetite ratings. In the study as a whole, preprandial
hunger ratings correlated positively, and fullness ratings correlated negatively, with food intake at the
subsequent meal. The lack of any stimulatory effect of
insulin on appetite ratings and food intake in this
study, in which hypoglycemia (and hence glucoprivation) was avoided, provides indirect evidence that the
increase in appetite and food intake that accompanies
insulin-induced hypoglycemia (13, 29) is due to glucoprivation.
Our findings are consistent with those of Woo et al.
(38), who administered intravenous insulin and glucose
infusions to young, nonobese, adult subjects, starting
30 min before and finishing 15 min after a test meal, for
a total duration of ,55 min. In a separate part of the
same study, subjects ate a small amount of food 12 min
before the test meal to induce endogenous insulin
release and then received an insulin-glucose infusion
for ,21 min from the beginning of the meal. Neither
insulin infusion affected appetite or food intake compared with control infusions. In that study, preprandial
plasma insulin concentrations were ,35–40 µU/ml,
lower than often observed after a mixed meal, and
blood glucose concentrations were not matched on the
control and treatment days in the first part of the study.
Downloaded from http://ajpregu.physiology.org/ by 10.220.32.247 on June 18, 2017
When the total infusion period (baseline to 170 min)
was analyzed, there was an effect of time on fullness
(F 5 125, P , 0.0001), with subjects rating themselves
as less full immediately before the meal and more full
after the meal than at baseline. There was no effect of
treatment type on fullness (F 5 0.35, P 5 0.8) and no
interaction between treatment type and time (F 5 0.7,
P 5 0.8). The same was so when only the period
between baseline and the start of the meal was analyzed (time, F 5 2.9, P 5 0.014; treatment, F 5 0.3, P 5
0.8; time 3 treatment, F 5 0.8, P 5 0.7). Notwithstanding the absence of a significant effect (as assessed by
ANOVA) of treatment type on ratings of hunger and
fullness, there was a trend toward greater fullness and
less hunger at 140 min (immediately before the start of
the meal) on the glucose-only infusion day than on any
of the other three treatment days (Fig. 5).
The durations of eating, weight, energy content, and
macronutrient composition of food eaten are shown in
Table 2. There was an effect of treatment type on
energy intake (F 5 4.1, P 5 0.015), which was significantly (,15%) lower on the glucose infusion day than
any of the other treatment days, none of which was
different from each other. There was no effect of study
order on energy intake in the first three study days,
when the treatments were randomized (1,096 6 101,
1,090 6 106, and 1,060 6 97 kcal, F 5 0.27, P 5 0.8).
EFFECT OF INSULIN ON HUMAN APPETITE AND FOOD INTAKE
mones is low in the fasting state, interactions occur
between insulin and the larger amounts of these hormones released by food, which were not evident during
the 2-h preprandial part of this study. Similarly, it is
possible that the presence of both hyperinsulinemia
and hyperglycemia may be required to affect food
intake; such an interaction could potentially explain
the reduction of food intake produced by infusion of
glucose only in the second part of the study. Therefore,
although the responses to the euglycemic insulin infusions suggest that hyperinsulinemia does not by itself
affect short-term feeding, they do not exclude the
possibility that insulin can affect feeding via an interaction with another postprandial factor. Nevertheless, we
would have anticipated that such interactions, if they
exist, should have been apparent after the start of the
meal. We observed that plasma insulin levels increased
to the same extent during meal ingestion on the control
and both insulin infusion treatment days, so that at
both the start and end of the meal, plasma insulin
levels were ,50 and 100 µU/ml higher on the low- and
high-dose insulin days, respectively, than on the control
day. Despite this, there was no difference in the duration of eating or the amount or composition of the food
eaten between any of the treatments, suggesting that
insulin has neither a direct role in terminating food
intake nor a dose-responsive interaction with other
hormones in producing this effect. The observed lack of
effect of insulin infusions on meal-induced insulin
release is consistent with recent reports that powerful
stimuli for insulin release, such as GLP-I, are able to
overcome the feedback effects of insulin on its own
release (1).
We have only addressed the effects of short-term
insulin administration in relatively low ‘‘physiological’’
doses. It is possible that insulin has effects on feeding
behavior that take longer than 2.5 h to manifest. The
liver has been postulated to be a source of satiety
signals (21). We did not measure portal insulin concentrations during the insulin infusions, but, at steady
state (after the first 20 min of the infusion), they are
likely to have been similar to peripheral concentrations, as the portal-peripheral insulin concentration
gradient is abolished by peripheral insulin infusion
(27). Assuming a portal to peripheral concentration
gradient of ,2.5 to 1 for endogenous insulin (2), the
portal insulin concentration during high-dose exogenous insulin infusion would have been about one-half
the normal postprandial peak. Therefore, it is possible
that feeding behavior would have been affected by
infusing more insulin to produce a higher, but still
physiological, portal insulin concentration. It is also
possible that higher portal insulin concentrations were
achieved during the glucose-only infusion of study 2
than the high-dose insulin infusion and that this could
account for the reduced food intake associated with the
glucose-only infusion. It should, however, be recognized
that the absence of a dose-responsive effect of the
150-min euglycemic insulin infusions, while not excluding an independent effect on short-term appetite and
food intake of physiological concentrations of insulin in
Downloaded from http://ajpregu.physiology.org/ by 10.220.32.247 on June 18, 2017
Short-term effects of insulin on feeding, particularly at
higher physiological postprandial plasma concentrations, therefore could not be totally excluded. We have
extended the findings of Woo et al. by demonstrating
that longer-duration (2.5 h) euglycemic insulin infusions, which produce plasma insulin concentrations as
high as those after a normal meal, also have no
short-term effect on appetite in humans. Intravenous
insulin infusions produce sustained increases in cerebrospinal fluid insulin concentrations within 30–45 min of
their commencement (37), so it is likely that the brain,
a proposed site of insulin effects on feeding (30), was
exposed to elevated insulin levels for most of the insulin
infusion period in this study.
It should be recognized that our study design did not
allow examination of several possible effects of insulin
on feeding behavior. First, the timing of food ingestion
was fixed, so that the subjects had no control over when
they ate. Insulin could potentially affect the timing of
spontaneous food ingestion. In a series of elegant
studies using animals and humans, free to eat when
they wished, Campfield et al. (5, 6) showed that the
timing of spontaneous food ingestion is related to
transient decreases in blood glucose concentration
within the nonhypoglycemic range, implying that a
reduction in glucose levels may trigger food intake. It is
conceivable that such decreases are related to increases
in insulin secretion or action, but this is probably
unlikely given that they occur in the fasting state, some
time after a meal. We are not aware of studies that have
addressed this issue.
Our study was designed specifically to evaluate the
effects of insulin and not to reproduce in every detail a
normal physiological state. The infusions were performed after an overnight fast, and it may be inappropriate to extrapolate the findings to times later in the
day when meals are spaced more closely. Nevertheless,
the duration of the overnight fast is not so great that
appetite ratings and food intake cannot be affected by
further changes within the physiological range. This is
indicated by the increasing ratings of hunger and
decreasing ratings of fullness during the two preprandial hours of the euglycemic infusions in this study, as
well as our previous observations, using a similar study
design, that appetite ratings and food intake in fasting
young adults are significantly suppressed by intraduodenal infusions of relatively modest amounts (340 kcal
over 2 h) of glucose (22) or lipid (9).
In healthy, lean individuals, an increase in plasma
insulin levels normally occurs in the postprandial
rather than the fasting state. Food ingestion stimulates
the secretion of a number of gastrointestinal hormones,
besides insulin, which may affect feeding. They include
glucagon (12), amylin (8), cholecystokinin (12, 20), and
GLP-I (16, 36). Animal studies suggest that the interaction of insulin with these hormones may have synergistic effects on appetite. For example, in nonhuman
primates, central administration of low-dose insulin
can enhance the satiating effects of peripherally administered cholecystokinin (11). It is theoretically possible
that, because secretion of these gastrointestinal hor-
R601
R602
EFFECT OF INSULIN ON HUMAN APPETITE AND FOOD INTAKE
this infusion were almost double those during the other
three treatment infusions (9.3 vs. 5.2–5.3 mmol/l). It
has been difficult to examine the effect of elevated blood
glucose concentrations on feeding, largely because of
the problems involved in manipulating blood glucose
levels while insulin levels and the amount of nutrient
administered are kept constant. Over 40 years ago,
Mayer proposed that food intake is modulated by
variations in blood glucose concentrations (24), but
increased emphasis has since been placed on the rate of
glucose utilization and the availability of glucose for
cellular oxidative processes (28). Although hepatic glucose production is suppressed during both intravenous
glucose infusions and euglycemic hyperinsulinemic
clamp studies, there is evidence that hyperglycemia
exerts a suppressive effect independent of and additional to that of hyperinsulinemia (10). Intravenous
infusion of glucose enhances splanchnic glucose uptake, whereas euglycemic infusion of insulin resulting
in plasma insulin levels up to 20 times higher does not.
The hyperglycemia-induced enhancement of splanchnic glucose uptake may be dependent on coexistent
hyperinsulinemia (10). Hyperglycemia is also associated with slowing of gastric emptying (23) and increased perception of fullness in both diabetic (19) and
nondiabetic subjects (17), potential mechanisms by
which it could exert suppressive effects on feeding.
Therefore, although most studies suggest that any
suppressive effect of hyperglycemia per se is likely to be
minor (22, 28, 40), one could speculate that the independent effects of hyperglycemia on glucose utilization
and/or its effects on gastrointestinal motility and perceptions could be responsible for the different effects on
appetite observed in the two parts of our study. Or, as
discussed, hyperinsulinemia and hyperglycemia may
interact to inhibit appetite and food intake, whereas
each factor alone is without effect.
Last, it is possible that a factor (or factors) cosecreted
with endogenous insulin in response to intravenous
glucose is responsible for the suppression of food intake
by intravenous glucose in this study. Many hormones
with possible satiating effects are secreted in response
to intravenous glucose, albeit in smaller amounts than
after oral glucose ingestion. They include GLP-I and
amylin; both have been shown to reduce food intake in
animal studies (8, 36), and GLP-I has recently been
reported to reduce appetite and food intake in humans
(16).
In summary, we have established that 2.5-h euglycemic insulin infusions, producing plasma insulin concentrations equivalent to those normally achieved after a
meal, have no effect on ratings of hunger and fullness or
voluntary food intake in young adults. This finding
suggests that, by itself, insulin does not affect shortterm feeding behavior and is consistent with a previous
report that physiological hyperinsulinemia is without
effect on short-term food intake (38). Long-term effects
of insulin on food intake, if proven to exist in humans,
are likely to be secondary to metabolic effects of insulin
and not insulin per se. The observation that intravenous glucose infusion results in a modest reduction in
Downloaded from http://ajpregu.physiology.org/ by 10.220.32.247 on June 18, 2017
the portal (and peripheral) circulations, provides no
evidence for such an effect.
In the second part of the study, glucose was infused
intravenously without insulin to match the amount and
timing of the glucose infusion on the high-dose insulin
day. During the glucose-only infusion, ratings of hunger
were decreased and ratings of fullness were increased
compared with the other three infusion days. Although
these differences were not statistically significant, they
are consistent with the suppressive effect of this infusion on food intake. Energy intake during the test meal
was ,15% less than during the control and insulin
infusions, despite the peak preprandial plasma insulin
concentration being essentially the same as on the
low-dose day and one-half that on the high-dose insulin
day. The absolute meal-induced increase in plasma
insulin concentrations was much greater on the glucoseonly infusion day than the other 3 days, even though
less food was eaten, due to the priming effect of prior
glucose exposure on pancreatic insulin release (7). In
animals, short-term intravenous glucose infusions apparently have little, if any, suppressive effect on food
intake (see Ref. 28 for review), and prolonged infusions
are required to produce marked suppression. In baboons, for example, there is a significant reduction in
food intake during the 14- to 21-day intravenous glucose infusions, but this suppressive effect takes several
days to develop and is not maximal until about day 10
(40). The effect of intravenous glucose on human appetite is less clear and has even been reported to increase
appetite and food intake (29). There are a number of
possible explanations for the reduced food intake observed with glucose infusion in this study.
By necessity, the treatment order of this study could
not be randomized, and an order effect cannot be
excluded. All subjects received the glucose-only infusion on their fourth and final study day to match the
infusion rate exactly to that on the high-dose insulin
day. The subjects may therefore have eaten less because they were being presented with the same meal
for the fourth time and were tired of it. However, the
lack of a significant order effect on food intake during
the first three randomized euglycemic studies perhaps
makes this explanation unlikely. The trend toward
increased fullness ratings and decreased hunger ratings preprandially on the glucose-only day may suggest
that subjects were experiencing reduced appetite even
before being presented with the meal. It is possible that
the effects of endogenous and exogenous insulin on food
intake differ. Such a difference has not been demonstrated but could occur as a result of the secretion of
endogenous insulin into the portal circulation and
passage through the liver before entering the peripheral circulation. Alternately, exogenous and endogenous insulin may exert the same qualitative effects on
feeding, and the suppression in this study reflects a
dose-responsive effect of portal insulin concentrations
on food intake, as discussed.
The hyperglycemia induced by glucose-only infusion
could potentially have mediated the associated suppression of food intake; blood glucose concentrations during
EFFECT OF INSULIN ON HUMAN APPETITE AND FOOD INTAKE
food intake may indicate the effect of hyperglycemia, of
another glucose-induced ‘‘postprandial’’ factor, or an
interaction between hyperinsulinemia and such a factor.
I. M. Chapman was supported by a Pharmacia Research Fellowship of the Royal Australasian College of Physicians.
Address for reprint requests: I. Chapman, Dept. of Medicine, Royal
Adelaide Hospital, North Terrace, Adelaide, South Australia, Australia 5000.
Received 5 June 1997; accepted in final form 29 October 1997.
REFERENCES
18. Holt, S. H., and J. B. Miller. Increased insulin responses to
ingested foods are associated with lessened satiety. Appetite 24:
43–54, 1995.
19. Jones, K. L., M. Horowitz, M. Berry, J. M. Wishart, and S.
Guha. The blood glucose concentration influences post-prandial
fullness in insulin dependent diabetes. Diabetes Care 20: 1141–
1146, 1997.
20. Kissileff, H. R., F. X. Pi-Sunyer, J. Thornton, and G. P.
Smith. C-terminal octapeptide of cholecystokinin decreases food
intake in man. Am. J. Clin. Nutr. 34: 154–160, 1981.
21. Langhans, W. Hepatic and intestinal handling of metabolites
during feeding in rats. Physiol. Behav. 49: 1203–1209, 1991.
22. Lavin, J., G. A. Wittert, W. M. Sun, M. Horowitz, J. E.
Morley, and N. W. Read. Appetite regulation by carbohydrates:
role of gastrointestinal hormones. Am. J. Physiol. 271 (Endocrinol. Metab. 34): E209–E214, 1996.
23. MacGregor, I. L., R. Gueller, H. D. Watts, and J. H. Meyer.
The effect of acute hyperglycaemia on gastric emptying in man.
Gastroenterology 70: 190–196, 1976.
24. Mayer, J. Glucostatic mechanism of regulation of food intake. N.
Engl. J. Med. 249: 13–16, 1953.
25. McIntyre, N., D. S. Turner, and C. D. Holdsworth. Intestinal
factors and insulin secretion (Abstract). Diabetologia 1: 73, 1965.
26. Panksepp, J., A. Pollack, K. Krost, R. Meeker, and M.
Ritter. Feeding in response to repeated protamine zinc insulin
injections. Physiol. Behav. 14: 487–493, 1975.
27. Peacey, S. R., A. Rostami-Hodjegan, E. George, G. T. Tucker,
and R. Heller. The use of tolbutamide-induced hypoglycemia to
examine the intraislet role of insulin in mediating glucagon
release in normal humans. J. Clin. Endocrinol. Metab. 82:
1458–1461, 1997.
28. Rezek, M. The role of insulin in the glucostatic control of food
intake. Can. J. Physiol. Pharmacol. 54: 650–665, 1976.
29. Rodin, J., J. Wack, E. Ferrannini, and R. A. DeFronzo.
Effect of insulin and glucose on feeding behaviour. Metabolism
34: 826–831, 1985.
30. Schwartz, M. W., D. P. Figlewicz, D. G. Baskin, S. C. Woods,
and D. Porte, Jr. Insulin in the brain: a hormonal regulator of
energy balance. Endocr. Rev. 13: 387–414, 1992.
31. Sepple, C. P., and N. W. Read. Gastrointestinal correlates of
the development of hunger in man. Appetite 13: 183–191, 1989.
32. Silverstone, T., and M. Besser. Insulin, blood sugar and
hunger. Postgrad. Med. J. Suppl. 47: 427–429, 1971.
33. Strubbe, J. H., and C. G. Mein. Increased feeding in response
to bilateral injection of insulin antibodies in the VMH. Physiol.
Behav. 19: 309–313, 1977.
34. Stunkard, A. J., and S. Messick. The three-factor eating
questionnaire to measure dietary restraint, disinhibition and
hunger. J. Psychosom. Res. 29: 71–83, 1985.
35. Thompson, D. A., and R. G. Campbell. Hunger in humans
induced by 2-deoxy-D-glucose: glucoprivic control of taste preference in food intake. Science 198: 1065–1068, 1977.
36. Turton, M. D., D. O’Shea, I. Gunn, S. A. Beak, C. M.
Edwards, K. Meeran, S. J. Choi, G. M. Taylor, M. M. Heath,
P. D. Lambert, J. P. Wilding, D. M. Smith, M. A. Ghatei, J.
Herbert, and S. R. Bloom. A role for glucagon-like peptide I in
the central regulation of feeding. Nature 379: 69–72, 1996.
37. Wallum, B. J., G. J. Taborsky, Jr., D. Porte, Jr., D. P.
Figlewicz, L. Jacobson, J. C. Beard, W. K. Beard, and D.
Dorsa. Cerebrospinal fluid insulin levels increase during intravenous insulin infusions in man. J. Clin. Endocrinol. Metab. 64:
190–194, 1987.
38. Woo, R., H. R. Kissileff, and F. X. Pi-Sunyer. Elevated
postprandial insulin levels do not induce satiety in normalweight humans. Am. J. Physiol. 247 (Regulatory Integrative
Comp. Physiol. 16): R745–R749, 1984.
39. Woods, S. C., L. J. Stein, L. D. McKay, and D. Porte, Jr.
Chronic intracerebroventricular infusion of insulin reduces food
intake and body weight in baboons. Nature 282: 503–505, 1979.
40. Woods, S. C., L. J. Stein, L. D. McKay, and D. Porte, Jr.
Suppression of food intake by intravenous nutrients and insulin
in the baboon. Am. J. Physiol. 247 (Regulatory Integrative Comp.
Physiol. 16): R393–R401, 1984.
Downloaded from http://ajpregu.physiology.org/ by 10.220.32.247 on June 18, 2017
1. Ahren, B., H. Larsson, and J. J. Holst. Effects of glucagon-like
peptide-I on islet function and insulin sensitivity in noninsulindependent diabetes mellitus. J. Clin. Endocrinol. Metab. 82:
473–478, 1997.
2. Blackard, W. G., and N. C. Nelson. Portal and peripheral vein
immunoreactive concentrations before and after glucose infusion. Diabetes 19: 302–306, 1970.
3. Booth, D. A., and T. Brookover. Hunger elicited in the rat by a
single injection of bovine crystalline insulin. Physiol. Behav. 3:
439–446, 1968.
4. Brief, D. J., and J. D. Davis. Reduction of food intake and body
weight by chronic intraventricular insulin infusion. Brain Res.
Bull. 12: 571–575, 1984.
5. Campfield, L. A., and F. J. Smith. Transient declines in blood
glucose signal meal initiation. Int. J. Obes. Suppl. 14: 15–33,
1990.
6. Campfield, L. A., F. J. Smith, M. Rosenbaum, and J. Hirsch.
Human eating: evidence for a physiological basis using a modified paradigm. Neurosci. Biobehav. Rev. 20: 133–137, 1996.
7. Cerasi, E. Potentiation of insulin release by glucose in man. Acta
Endocrinol. 79: 511–534, 1975.
8. Chance, W. T., A. Balasubramaniam, A. Stallion, and J. E.
Fischer. Anorexia following the systemic injection of amylin.
Brain Res. 607: 185–188, 1993.
9. Cook, C. G., J. M. Andrews, K. L. Jones, G. A. Wittert, I. M.
Chapman, J. E. Morley, and M. Horowitz. The effects of small
intestinal nutrient infusion on appetite and pyloric motility are
modified by age. Am. J. Physiol. 273 (Regulatory Integrative
Comp. Physiol. 42): R755–R761, 1997.
10. DeFronzo, R. A., E. Ferrannini, R. Hendler, P. Felig, and J.
Wahren. Regulation of splanchnic and peripheral glucose uptake by insulin and hyperglycemia in man. Diabetes 32: 35–45,
1983.
11. Figlewicz, D. P., L. J. Stein, D. West, D. Porte, Jr., and S. C.
Woods. Intracisternal insulin alters sensitivity to CCK-induced
meal suppression in baboons. Am. J. Physiol. 250 (Regulatory
Integrative Comp. Physiol. 19): R856–R860, 1986.
12. Geary, N., H. R. Kissileff, F. X. Pi-Sunyer, and V. Hinton.
Individual, but not simultaneous, glucagon and cholecystokinin
infusions inhibit feeding in men. Am. J. Physiol. 262 (Regulatory
Integrative Comp. Physiol. 31): R975–R980, 1992.
13. Geiselman, P., and D. Novin. Sugar infusions can enhance
feeding. Science 218: 409–491, 1982.
14. Genuth, S. M. Plasma insulin and glucose profiles in normal,
obese, and diabetic persons. Ann. Intern. Med. 79: 812–822,
1973.
15. Grossman, M., and I. F. Stein. Vagotomy and hungerproducing action of insulin in man. J. Appl. Physiol. 1: 263–269,
1948.
16. Gutzwiller, J. P., B. Goke, J. Drewer, S. Ketterer, D.
Handschin, P. Hilldebrand, R. Winterhalder, D. Conen,
and C. Beglinger. Glucagon-like peptide-I is a physiologic
regulator of food intake in humans (Abstract). Gastroenterology
112: A1153, 1997.
17. Hebbard, G. S., M. Samsom, J. M. Andrews, D. Carman, B.
Tansell, W.-M. Sun, J. Dent, and M. Horowitz. Hyperglycemia affects gastric electrical rhythm and nausea during intraduodenal triglyceride infusion. Dig. Dis. Sci. 42: 568–575, 1997.
R603