- Email: [email protected]
Reflex insulin response associated to food intake in human subjects
Physiology & Behavior, Vol. 31, pp. 515--521. Pergamon Press Ltd., 1983. Printed in the U.S.A.
Reflex Insulin Response Associated to Food Intake in Human Subjects F. B E L L I S L E , J. L O U I S - S Y L V E S T R E , F. D E M O Z A Y , D. B L A Z Y A N D J. L E M A G N E N
Laboratoire de Neurophysiologie Sensorielle et Comportementale, CollOge de France 11 Place Marcelin Berthelot, 75231 P A R I S Cedex 05, France R e c e i v e d 17 J a n u a r y 1983 BELLISLE, F., J. LOUIS-SYLVESTRE, F. DEMOZAY, D. BLAZY AND J. LE MAGNEN. Reflex insulin response associated to food intake in human subjects. PHYSIOL BEHAV 31(4) 515-521, 1983.--The occurrence of a reflex insulin discharge at the beginning of a meal, and its possible influence on intake were studied in 7 normal weight humans. Each subject was tested twice under three standard meal conditions. The evolutions of insulinemia and glycemia were recorded over an 84 min observation period, starting 2 min before food presentation. Blood was drawn continuously from an antecubital vein, and collected in l-min samples for the first 30 min, and then in 3-min samples. The average glycemia curve was stable until some 18-20 min after meal onset. By contrast, a significant rise in plasma insulin appeared as early as the 4th min after meal onset and it is hypothesized to be preabsorptive, of cephalic and/or gastric origin. However, inter-test variations were large even in the same person. Schematically, three types of early insulin responses were observed: high and/or sustained rise, moderate and/or short increase, moderate decrease in l~lasma insulin. The shape of the early insulin response was not related to any meal characteristic. The potential biological and behavioral significance of the early insulin release is discussed. Insulin
Glycemia
Pre-absorptive reflexes
Cephalic phase
IN rats, an early preabsorptive release of insulin is observed within the first minutes following the intake of food [10,13]. This reflex response is biphasic: a first peak is of cephalic origin and appears within 3 minutes; a second peak follows between 3 and 6 minutes, and is likely elicited at the gastro-intestinal level. The amplitude of the insulin discharge appears to be dependent on the sensory qualities of the food presented: a more preferred food induces a larger cephalic insulin release than a low palatability food; furthermore, since this early insulin response disappears long before the end of the meal, it has been suggested that the amount of food eaten could be influenced by the taste-modulated insulin release [12]. In rats habituated to scheduled meals, the cephalic peak of the reflex insulin response can be conditioned to the mere appearance of the food cup [12] or to an arbitrary signal (time of day, or odor) that reliably predicts the imminence of food presentation [23,24]. A cholinergic mechanism has been shown to mediate the response as it is eliminated by atropine [24], and the determining role of the vagus has been ascertained by the disappearance of the cephalic response after vagotomy [10]. It has also been reported that the preabsorptive release of insulin facilitates and accelerates the post-ingestive glucose uptake in rats [ll]. In dogs, such a reflex insulin discharge has been observed within 10 minutes after an oral glucose load, while the plasma glucose level remained unchanged [5]. However satisfactory, this clear and coherent picture emerged from averaged insulinemia and glycemia curves which masked very large individual differences in the early insulin response to oral glucose. It has also been shown that dogs respond by
Food intake
Human subjects
a reflex insulin discharge to sham feeding of glucose, to tap water or cellulose presented orally [5, 6, 9]. In humans, a reflex insulin release of cephalic origin has been reported after visual and/or olfactory presentation of palatable foods in both lean and obese individuals [15, 18, 20], or even after hypnotic suggestion of food [8]. In these studies, the subjects were tested on a single, uniform stimulation condition. In spite of this, the insulin response often appeared to vary in amplitude and time course in different subjects. The most detailed study of this phenomenon in humans [20] included both lean and obese women and revealed that during visual and olfactive stimulation by food, the plasma insulin level could be highly elevated, somewhat elevated, or somewhat decreased relative to the level observed prior to food presentation. The proportions of the tests representing high, low, or negative cephalic insulin responses were not specified. In the same study, replication was attempted in obese subjects and the reflex insulin response exhibited poor within-subject reproducibility. In a sub-group of obese women, no change in the plasma insulin concentration were noted up to 20 rain after the visual presentation of a non food stimulus; this control was not reported for lean subjects. It is thus difficult to establish how frequently or reliably a given type of response is likely to appear in a normal population, or whether the amplitude of the response is a function of some subject's characteristic, or of the particular palatability conditions for a given individual. Since the particular type of food used in the experiment might have been differently palatable for different persons, it can be suggested that perhaps part of the variation in the cephalic response was due to differences in palatability. Be-
Copyright © 1983 Pergamon Press Ltd.--0031-9384/83/100515-07503.00
sir
!~t~[ ! iSl,l i : : :
sides, the relevance of such a reflex response for actual food consumption remains unknown. Obviously, studies of reflex, conditioned insulin release in humans present considerable technical difficulties since they involve blood sampling at short intervals, before, during and shortly after food presentation. Many factors associated with the procedure are likely to interfere with and modify the insulin level, stress being the most obvious. In all of the previous human studies, blood was collected at discrete intervals (1-10 min) shortly before, during, and shortly after the presentation of food stimuli and the reiteration of the blood sampling procedure may have acted as a factor of renewed tension or stress. Furthermore, the fact that the subjects were presented with food that they were not allowed to eat may have modified the cephalic response, as it alters the salivary response [25]. The occurrence of a preabsorptive, reflex insulin response has never been studied during actual consumption in human subjects. The reflex response which occurs in rats and humans on cephalic stimulation by food stimuli is likely to occur at the beginning of a meal. It might perhaps reflect palatability, and exert some influence on eating behavior. In the present experiment, the variations in insulinemia associated to food intake were examined in order to quantify such an eventual early insulin response in humans and to study its relationships with behavioral or metabolic events associated to the meal. Many studies in humans have examined the evolution of insulinemia and glycemia following different types of meals. Although these studies are highly informative about the differential absorption of certain types of foods, they use rather long blood sampling intervals (> 15 rain) that do not allow the observation of early reflex phenomena [2, 14, 16, 22, 26]. In the present experiment, seven human subjects were presented with many standard sandwich meals of different palatability. Their blood was collected continuously during and after the meal. Plasma insulin and glucose were assayed so as to obtain a fine-grain picture of their evolution from meal onset until 82 min later. The shape of the insulin curve in the first minutes of the meal, particularly before the glycemia curve started to reflect absorption, was followed closely. The amplitude of any early insulin variations and their relationships with palatability conditions, eating behavior, and the metabolic consequences of the meal were examined. Since subjects were tested repeatedly, the reproducibility of their early insulin responses was assessed. The experiment was conducted at Hopital Bichat (Paris) and Hopital du Kremlin-Bicetre. We would like to thank Dr. M. Apfelbaum and Dr. G. Barres for their cooperation. METHOD
Subjects The subjects were 4 women and 3 men, between 20 and 25 years of age. Their body weight was within -+ 10% of the ideal weight, according to the Metropolitan Life Insurance Co. Tables (1959). They were in good health at the time of the experiment. All were medical students who normally ate lunch at the hospital cafeteria. Therefore a sandwich lunch was not atypical for them. Before the beginning of the experiment, their alimentary habits and their attitudes towards food were assessed. The subjects had never been underweight or overweight in the past and had never had to diet. They exhibited no alimentary restraint; they habitually ate until complete satiation and never worried about their weight.
M~'ol (':,/;difion',
Three different meal conditions were presented ~wicc i~; each subject (total::6 meals per sutLjectL All meals were composed of standard, cocktail size sandwiche> These sandwiches, or food units, consisted of a 3 cm ~ piece ot commercial white bread, spread with a thin, uniform layer of a food of distinctive aspect and flavor. The nulritiomd value and texture of these food units, being determined largely hy the bread, were approximately equal lbr all typc~ of ~md wiches. (They represented approximately lt¢~ KJ :rod weighed about 7 grams.) In a mixed meal condition, five different spreads were presented: crab, anchovies, liver pate, pork pate and bu|ter. The other two meal conditions were high and low preference conditions, in which only the most preferred, ¢~ the lease preferred food was presented. The preference scale wa~ e~.tablished for each subject prior to the series of [ests, 013 the basis of a visual analogue test, as described ir~ a previous paper [ll.
Blood Sampling From an indwelling catheter (Vygon, 121 13, France) implanted in the subject's antecubital vein, blood was drawn in a continuous fashion by means of a peristaltic pump, at a flow rate of 0.8 ml per min, while a heparin solution (2000 tzl/ml saline) was delivered inside the catheter through a Butterfly needle (Abbott, 21-INT, Ireland) at a flow rate of 0.04 ml per min, so as to prevent clotting. Blood samples were collected minute per minute for the first 30 rain and then over 3 min intervals, up to a lotal of 84 min. They were centrifuged within a few minutes after drawing. Plasma was collected and stored at 24°C. Plasma insulin was measured in duplicate with the INSIK-1 radioimmunoassay kit (CEA, France). The interassay coefficient of variation was 12%. Plasma samples were assayed for glucose concentration with a YSI Glucose Analyser (YSI, 23A, Bioblock, France). The standard error with this method was + 1%.
Procedure The experimental sessions took place in quiet hospital rooms. The subjects were always tested alone. On test days, the subjects had their regular breakfast at the usual time, and were instructed to avoid any additional eating or unaccustomed physical activity before the test. The level of hunger was thus moderate at the beginning of the experimental sessions, and was expected to be representative of the normal deprivation conditions at lunch time. Each subject came to the experimental settings once 'a week, at 10:30 a.m. The subject was asked to sit in a comfortable armchair and the catheter was implanted immediately. It was kept patent through a saline perfusion (0.85%) which was continued for at least one hour before the beginning of blood drawing. This was done so that the effects of any stress due to the implantation of the catheter have disappeared by the time of blood drawing. No anaesthetics were used and none of the subjects showed or reported any mood perturbation at the time of catheterization. The subjects were then left to rest until the beginning of blood drawing, approximately 1 hour later. Immediately before blood drawing, the tip of the heparin delivering needle was inserted inside the catheter, and a constant heparin inflow to the catheter was started. The peristaltic pump was
R E F L E X I N S U L I N RESPONSE AND E A T I N G
517
then put into action to draw the heparinized blood. This represented no other puncture of the subject's arm, and the whole procedure was apparently tolerated without stress. Blood drawing was started at about 12:00 p.m. The blood sampled during the first 5-10 min was discarded. After the Experimenters felt sure that the blood and heparin flows were stable, the first 1 min blood sample was collected. The test food was prepared in an adjacent room a few minutes before being served, and was presented on a tray during the third minute of blood collection. The subjects were instructed to eat as much or as little as they wished. The duration and size of the meal were thus freely determined by the subjects. Prandial drinking was allowed ad lib. After the subject had stopped eating, the remaining food was removed. Meal size (number of sandwiches ingested) and meal duration were noted. Eating rate (number of sandwiches ingested per minute) was evaluated, since in a previous study [1] it was shown to be a very sensitive index of palatability. For one trial no food was presented, so as to measure the spontaneous variations of insulinemia and glycemia. The order of the 7 different trials was randomly determined. The Experimenters remained in the room during the whole session, including the meal.
/
linsuline
12108" i
i
i
i
I
I
i
.
1 2 3 4 5 6 "1 8 ; 10111' I~Z 1; 14 1!i 16 17 18 1~)2om|n FIG. i. Spontaneous variations of insulinemia and glycemia during 20 minutes under standard test conditions but without food presentation (means of seven tests_+SEM).
RESULTS Meal size and eating rate were compared between the three meal conditions. In the low preference condition 32_+4.1 sandwiches were ingested, and 37.5_+3.8 in the high preference condition. A two-way analysis of variance with repeated measurements revealed that these figures were not statistically different, F(1,6)=2.44, n.s. During mixed meals, the subjects ingested 44.9_+4.3 sandwiches. This is signifi-
Statistics
Analyses of variance and correlations were computed according to standard formulas [4]. Means_+SEM are presented in the text and figures.
120-
.glucose mg/m I
GLUCOSE mg/dl T
11010090-
80
NSULINE
-
-
",.~'/."
40-
30-
.
~
10-1 ~J
~,
lb
l~S
z,b
z's
3b
d.
~
~s
do
.g
~o
~
Yo
Ys
do M~N
FIG. 2. Variations in insulinemiaand glycemia during an 84 minute peri-prandial observation period (average of 42 meals +_SEM). Food was presented to the subjects during the 3rd minute of blood collection. The stars indicate the first sample significantly different from baseline (first 2 samples). While no significant change in the glucose level occurs before the 17th minute after the beginningof the meal, an early rise in plasma insulin becomes significant from the 4th minute after meal onset. After a pre-absorptive peak (Sth minute), plasma insulin decreases, but remains significantly elevated as compared to pre-meal values. After the 17th minute, insulinemiaand glycemia are closely parallel: they rise rapidly, reach a peal: level between 50 and 60 minutes after meal onset, after which they decrease slowly.
t~:l.I tSl.t: /: t ~i
518 .~
iNSULINE i~ U / m l
25
12 11
20
10 9 8
15-
7 6
10-
5 4 3
5-
2 1 0
~-.~..~_
__................................................................
o[-5O
-2'5
8
is
50
7'5
16o
I NSULINE
pU
-1
FIG. 4. Histogram presenting the distribution in amplitude of the cephalic insulin responses observed during 42 meals. The surface area of the insulinemia curve above or below baseline was integrated over the first 16 minutes after meal onset, before the appearance of any sign of absorption.
4 3 2 1
0 -1 -2
1 0
.....
1 -2
-1
o
t
2
3
4
5
6
7
s
9
lO
11 1 2 13 14 t 5
18
minutes
FIG. 3. Three types of preabsorptive insulin responses to the ingestion of food, as compared to baseline values. c antly more than in the low, F( 1,6) = 25.33, p < 0.01, and high, F(1,6)=6.5, p<0.05, preference conditions. Meal duration was 8.33_+3.5 minutes in the low preference condition and 10.7_+4.8 minutes in the hi_g_hpreference condition; these values are not significantly different, F(1,6)-=5.67, n.s. iVfixed meals lasted 11.8___3.8 minutes, which is significantly longer than low preference meals, F(1,6)=7.19, p<0.05, but not than high preference meals, F(1,6)=2.39, n.s. Eating rate (number of sandwiches ingested per minute) was identical in the three meal conditions: 3.9_+0.3 (low preference), 3.7_+0.2 (high preference) and 3.8_+0.2 (mixed meal): F(2,12)=0.31, n.s.
Figure 1 presents the spontaneous variations in insulinemia and glycemia at the hour of the test, without food presentation. Glycemia was extremely stable. While insulinemia was more variable, t tests for differences between correlated samples revealed that the two extreme points (3rd and 5th minutes) were not significantly different from each other, two-tailed t(6)=2.03, n.s. Figure 2 presents the evolution of insulinemia and glycemia associated to food intake (84 min observation period). These curves represent the average of the 42 meals. The glycemia graph is perfectly stable until the 17th minute; then a rapid increase occurs and glycemia remains elevated until the end of the observation period. On the basis of this observation, the continuous insulin and glucose curves were analyzed in two phases. The data gathered from the beginning to the 16th min of the meal were regarded as a first early phase, during which reflex fluctuations of cephalic or gastric origin might be observed. Data from the 17th min, up to the
82nd after meal onset, were regarded as a second phase, primarily describing absorption. T tests for difference between correlated samples revealed a significant elevation of the insulin level over baseline (average of the two minutes preceding food presentation), during the early phase. Insulinemia was elevated from the 4th min of the meal, two-tailed t(40) = - 3.13, p <0.01. On the 8th min after meal onset, insulinemia reached a peak value which was significantly elevated as compared to both baseline, two-tailed t(40)=5.64, p<0.001, and to the lowest point reached between this peak and the beginning of the absorptive phase, on the 12th rain after meal onset, twotailed t(40)=2.44, p <0.05. In turn, this last insulin value was significantly higher than baseline, two-tailed t(40)=3.62. p<0.001. An examination of individual curves revealed large differences in the insulin response. In the 42 experimental sessions, three main types of early insulin responses were observed: high and/or sustained elevation above baseline, low amplitude increase, and low amplitude decrease relative to baseline. The graphs presented in Fig. 3 illustrate these three early insulin patterns. A representative example is given of each type of response. Insulinemia and glycemia were analyzed in terms of the surface area above or below baseline (data observed prior to food presentation) for a given observation period (prior to or posterior to the rise in glycemia). The surface area of the early insulin response was computed for all 42 meals. The surface areas of insulinemia and glycemia beyond the 16th min of the meal were computed for all the tests in which blood sampling was successfully carried over for 84 min (N=34). Occlusion of the IV catheter prevented the collection of a complete series of data points in the other meals (N =8). Figure 4 shows a histogram of the amplitudes of the cephalic insulin responses, as integrated over 16 min. It appears that in only 5 out of the 42 meals (11.9%), the insulin curve was negative relative to baseline. In all other cases, the insulin curve was over baseline and the histogram reveals a continuous distribution of amplitudes, almost normal in shape. When no food was presented, the surface area of insulin variation around baseline, as integrated over 16 rain. was + 1 . 3 / z U (average of 7 tests). The profile of the cephalic insulin response was not linked in any simple fashion to subjects' characteristics or to meal
REFLEX INSULIN RESPONSE AND EATING composition: a given subject could exhibit largely different preabsorptive insulin responses, when presented repeatedly with the same food. During mixed meals, the surface area of the early insulin response was +31.5-+9.9/~U; in the high preference condition, it was +48-+9.1/zU; in the low preference condition, it was +50.7_+ 14.9/xU. A two way analysis of variance with repeated measurements indicated that these values were not significantly different: F(2,12)=0.81, n.s. The amplitude of the early insulin response was not correlated to meal size (number of food pieces ingested): r ( 4 0 ) = - 0 . 0 6 , n.s. ; to meal duration: r(40)=-0.188, n.s.; or to eating rate: r(40)=0.257, n.s. The positive early insulin response was usually perceptible for many minutes. Its time course was highly variable across different meals, even in the same subject. As shown in Fig. 3, one or two peaks could be observed before the sustained rise due to absorption. The evolutions of insulinemia and glycemia following the beginning o f absorption were different between test meals. Not only did the total surface area of insulin and glucose vary between meals, but the time course of the responses was variable. Generally, a sharp increase in plasma insulin and glucose was noted around the 16--18th min. In most cases, a high plateau was reached and the plasma concentrations started to decrease before the 84th min. In other cases, the increase in plasma insulin and glucose concentrations was slower to appear and/or no decrease was noted at the end of the observation period. The highest levels of insulinemia and glycemia were reached at unpredictable intervals after meal onset. The surface area of the early insulin response (first 16 min) was not found correlated to the surface area of the absorptive (17th-82nd min) insulinemia, r(32)=0.034, n.s., or glycemia, r(32)=0.118, n.s. The absorptive insulin surface area was not correlated to absorptive glycemia, r(32)=0.151, n.s.
Meal size was positively correlated to the amplitude of the integrated insulin response during the absorptive phase, r(32)=0.425, p <0.02, but not to the surface area of glycemia during the same period, r(32)=-0.088, n.s. DISCUSSION
An early phase of insulin release can be observed at the beginning of food intake in human subjects. In contrast with the stable insulinemia observed in the absence of food, an increase in plasma insulin, relative to baseline, was usually recorded within the first 16 min of food consumption, while the plasma glucose level remained unchanged. The total prandial and postprandial insulinemia curve therefore appeared to be clearly biphasic. In a first phase, which was contemporary with a stable glucose level, a variation in insulin level occurred a few minutes after the beginning of eating; insulinemia then receded towards baseline values, at about 10-12 min after meal onset. In a second phase, a sharp rise in insulinemia occurred concurrently with an increase in glycemia; this elevation was much more important than that of the first phase, and lasted over 60 min. This is consistent with prandial and post-ingestive insulinemia curves obtained in rats [10, I1, 21] and dogs [5, 6, 7]. In particular, the shape of the insulinemia curves presented here is very similar to that observed in dogs after an oral glucose load [5]. In these animals, one or two peaks of insulin could be observed before the rise in plasma glucose reflecting absorption: this also appeared in our human subjects as shown in the positive insulin responses presented in Fig. 3. Like our human sub-
519 jects, the dogs exhibited very irregular individual responses, even though they were simply presented with glucose. These observations suggest that the insulin variations which preceded the elevation in glycemia in our human subjects were preabsorptive and involved a reflex mechanism, which could be of cephalic and/or gastrointestinal origin. Of course, the early absorption of some nutrients before any variation was detectable in the glucose level cannot be completely ruled out. H o w e v e r the occurrence of similar insulin responses in the absence of ingestion or in sham feeding in various species [5, 8, 9, 11, 15, 18, 20] suggests the idea that the early insulin response observed here mainly reflects the action of reflex mechanisms in which cephalic stimulation could play a determinant role. Furthermore, the fact that the early increase in insulin is transient and recedes significantly before the clear concurrent elevation o f insulin and glucose is more easily understood if the early insulin response is considered the result of a preabsorptive reflex than that of a precocious absorption of nutrients by as yet unknown effectors: if the early insulin response reflected the beginning of absorption, it is difficult to see why insulinemia should exhibit one or two peaks, and then recede towards baseline values before a final sustained elevation. In spite of important differences in method (actual intake vs. visual stimulation; sandwich meal vs. varied hot meal), the early response reported here is extremely close to the reflex discharges observed in other human studies. Sjostrrm et al. [20] found comparable alterations in plasma insulin to occur during 5 min of actual visual and olfactory stimulation and to disappear on removal of the food stimulus. ParraCovarrubias et al. [15], working in obese adolescents, found a very high amplitude cephalic insulin response to outlast 15 min of visual and olfactory food presentation. The responses observed by Sjostrrm et al. [20] were almost immediate, appearing clearly within 1 or 2 min, while those reported by Parra-Covarrubias developed more slowly, and were still very modest after 5 min of stimulation. In the present study, the early insulin responses developed in a gradual manner over the first 8 min of food consumption and often extended beyond the duration of the meal. Differences in protocol could perhaps explain these different results. The deprivation time was longer in the Sjost r r m et al. [20] study and the protocol requested the subjects to skip a meal (breakfast) on test days. Although the obese subjects of Parra-Covarrubias et al. [15] were fasted overnight (10-12 hours), the experimental test was scheduled at normal breakfast time. Our subjects were only fasted for about 4 hours (between breakfast and lunch). Thus a more intense actual and/or perceived deprivation state might have led to steeper, more immediate cephalic responding in the Swedish study. This suggests that deprivation could exert a critical influence on the time-course of the cephalic responses, if not necessarily on their amplitude. The factors that command the amplitude or the shape of the reflex insulin response remain to be specified. The shape of the cephalic insulin curve does not appear to be determined in any direct fashion by stable subjects' characteristics, since a subject repeatedly presented with the same sandwich meal could exhibit largely different early insulin responses. In particular it does not appear that the subjects' " e x t e r n a l i t y " [17], or their relative sensitivity to food stimuli, could play a decisive role in determining the insulin secretion curve. On the other hand, immediate affective or cognitive alterations could partly determine the response. The amplitude of the cephalic insulin response was
520
BELI~ISI.I,I L7 Al
neither predicted by, nor predictive of any behavioral or physiological parameter, as measured in the present study. In particular, unlike what has been observed in rats by Louis-Sylvestre and Le Magnen [12], the amplitude of the integrated early insulin release did not vary coherently with palatability and was not found correlated to meal size. Two observations might help to explain these discordant results. First, it is clear that the palatability range in the two experiments was very different; in the study on rats, ingestion of a quinine-adulterated food was compared to that of a saccharin-edulcorated one; in humans, two unadulterated commercial foods were presented in sandwich form to the subjects. Contrary to what was observed in a previous work using a similar protocol [1], neither meal size nor eating rate, two sensitive indices of palatability, were significantly different between the two single flavor meals conditions in the present study, which indicates that probably the difference in palatability was extremely modest. This palatability range might not have been wide enough to produce important and systematic variations of the reflex insulin response. Secondly, it should be noted that in rats, the preabsorptive insulin response is of very short duration (less than 3 minutes) and has terminated before the end of the meal whereas in the present experiment the sandwich meals lasted for 10.3_+1 min on the average, and often terminated before the rise in glycemia signalling intestinal absorption. Even though most eating behavior was contemporary with the preabsorptive
physiological responses, no parameter of the meal wza~, shown to vary with the amplitude of the cephalic insulin response. Perhaps a longer meal, more varied in its com~ position, could ultimately be influenced by the shape of the early insulin response, although the mechanism of such a delayed effect of an early metabolic response on subsequent intake is difficult to imagine. As shown in Fig. 2, the early insulin response is extremely small in amplitude as compared to the post. absorptive secretion of insulin. In rats, the preabsorptive insulin response is proportionally greater, even in the case of a low palatability food [12]. The biological significance of such a weak response in humans, particularly its value as an anticipatory response to the absorption of nutrients, remains open to question, as it does not appear to reflect the level of palatability and predicts neither meal size, nor the future evolutions of insulinemia and glycemia. In humans as well as in animals, the prandial, preabsorptire insulin secretion response is probably only a part of the total neuroendocrine anticipatory response to the meal. For example, a cephalic phase of glucagon secretion [3~191 probably interacts with the insulin response to determine the subsequent physiological and behavioral responses. A more complete measurement of the total reflex response to food presentation might clarify its eventual impact on the ingestion and the absorption of a meal.
REFERENCES
1. Bellisle, F. and J. Le Magnen. The structure of meals in humans: Eating and drinking patterns in lean and obese subjects. Physiol Behav 27: 649-658, 1981. 2. Crapo, P. A., G. Reaven and J. Olefsky. Postprandial plasmaglucose and insulin responses to orally administered simple and complex carbohydrates. Diabetes 25: 741-747, 1976. 3. De Jong, A., J. H. Strubbe and A. B. Steffens. Hypothalamic influence on insulin and glucagon release in the rat. Am J Physiol 233: E380-E388, 1977. 4. Ferguson, G. A. Statistical Analysis in Psychology and Education. New York: McGraw Hill, 1971. 5. Fischer, U., H. Hommel, H. Fiedler and H. Bibergeil. Reflex mechanism on insulin secretion. Endocrinol Exp 8: 137-145, 1974. 6. Fischer, U., H. Hommel, M. Ziegler and E. Jutzi. The mechanism of insulin secretion after oral glucose administration. III Investigations on the mechanism of a reflectoric insulin mobilization after oral stimulation. Diabetologia 8: 385-390, 1972. 7. Fischer, U., H. Hommei, M. Ziegler and R. Michael. The mechanism of insulin secretion after oral glucose administration. I Multiphasic course of insulin mobilization after oral administration of glucose in conscious dogs. Differences to the behavior after intravenous administration. Diabetologia 8: 104-110, 1972. 8. Goldfine, I. D., C. Abraira, D. Gruenwald and M. S. Goldstein. Plasma insulin levels during imaginary food ingestion under hypnosis. Proc Soc Exp Biol Med 133: 274-276, 1970. 9. Hommel, H., U. Fischer, K. Retzlaff and H. Knofler. The mechanism of insulin secretion after oral glucose administration. II Reflex insulin secretion in conscious dogs bearing fistulas of the digestive tract by sham-feeding of glucose or tap water. Diabetologia 8:111-116, 1972. 10. Louis-Sylvestre, J. Preabsorptive insulin release and hypoglycemia in rats. A m J Physiol 230: 56--60, 1976. 11. Louis-Sylvestre, J. Relationship between two stages of prandial insulin release in rats. Am J Physiol 235: E103-EI 11, 1978.
12. Louis-Sylvestre, J. and J. Le Magnen. Palatabability and preabsorptive insulin release. Neurosci Biobehav Rev 4: Suppl I, 43-46, 1980. 13. Louis-Sylvestre, J. and J. Le Magnen. Phase cephalique de secr6tion d'insuline et vari6te des aliments au cours du repas chez le rat. Reprod Nutr Ddv, in press. 14. O'Dea, K., P. J. Nestel and L. Antonoff. Physical factors influencing postprandial glucose and insulin responses to starch. Am J Clin Nutr 33: 760-765, 1980. 15. Parra-Covarrubias, A., I. Rivera-Rodriguez and A. AlmarazUgalde. Cephalic phase of insulin secretion in obese adolescents. Diabetes 20: 800-802, 1971. 16. Potter, J. G., K. P. Coffman, R. L. Reid, J. M. Krall and M. J. Albrink. Effect of test meals of varying dietary fiber content on plasma insulin and glucose response. Am J Clin Nutr 34: 328334, 1981. 17. Rodin, J. The role of perception of internal and external signals on the regulation of feeding in overweight and non-obese individuals. In: Appetite and Food Intake, Lift, Sciences Report 2, edited by T. Silverstone. Berlin: Dahlem Konferenzen, 1976, pp. 265-283. 18. Sahakian, B. J., M. E. Lean, T. W. Robbins and W. P. T. James. Salivation and insulin secretion in response to food in non-obese men and women. Appetite 2: 209-216, 1981. 19. Samoles, E., J. Tyler, V. P. Rege and V. Marks. The Possible Role of Glucagon (Pancreatic and Extrapancreatic) in Insulin Secretion. Stockholm: VIth Congress International Diabetes Federation, 1967, pp. 446-454. 20. Sjostr6m, L., G. Garellick, M. Krotkiewski and A. Luyckx. Peripheral insulin in response to the sight and smell of food. Metabolism 29: 901-909, 1980. 21. Steffens, A. B. Plasma insulin content in relation to blood glucose level and meal pattern in the normal and hypothalamic hyperphagic rat. Physiol Behav 5: 147-151, 1970.
REFLEX
INSULIN RESPONSE
AND EATING
22. Vaaler, S., K. F. Hanssen and O. Aagenaes. Plasma glucose and insulin responses to orally administered carbohydrate-rich foodstuffs. Nutr Metab 24: 168-175, 1980. 23. Weiley, J. H. and G. A. Leveille. Significance of insulin in the metabolic adaptation of rats to meal ingestion. J Nutr 100: 1073-1080, 1970. 24. Woods, S. C., J. R. Vasselli, E. Kaestner, G. A. Szakmary, P. Milburn and M. V. Vitiello. Conditioned insulin secretion and meal feeding in rats. J Comp Physiol Psycho191: 128--133, 1977.
521
25. Wooley, S. C. and O. W. Wolley. Salivation to the sight and thought of food: a new measure of appetite. Psychosom Med 35: 136--142, 1973. 26. Zaccaria, M., E. De Palo, E. Zago, N. Sicolo, G. Erie and G. Federspil. Metabolic and endocrine responses to a standard mixed meal. A physiologic study. Acta Diabetol Lat 16: 45-53, 1979.
Reflex Insulin Response Associated to Food Intake in Human Subjects F. B E L L I S L E , J. L O U I S - S Y L V E S T R E , F. D E M O Z A Y , D. B L A Z Y A N D J. L E M A G N E N
Laboratoire de Neurophysiologie Sensorielle et Comportementale, CollOge de France 11 Place Marcelin Berthelot, 75231 P A R I S Cedex 05, France R e c e i v e d 17 J a n u a r y 1983 BELLISLE, F., J. LOUIS-SYLVESTRE, F. DEMOZAY, D. BLAZY AND J. LE MAGNEN. Reflex insulin response associated to food intake in human subjects. PHYSIOL BEHAV 31(4) 515-521, 1983.--The occurrence of a reflex insulin discharge at the beginning of a meal, and its possible influence on intake were studied in 7 normal weight humans. Each subject was tested twice under three standard meal conditions. The evolutions of insulinemia and glycemia were recorded over an 84 min observation period, starting 2 min before food presentation. Blood was drawn continuously from an antecubital vein, and collected in l-min samples for the first 30 min, and then in 3-min samples. The average glycemia curve was stable until some 18-20 min after meal onset. By contrast, a significant rise in plasma insulin appeared as early as the 4th min after meal onset and it is hypothesized to be preabsorptive, of cephalic and/or gastric origin. However, inter-test variations were large even in the same person. Schematically, three types of early insulin responses were observed: high and/or sustained rise, moderate and/or short increase, moderate decrease in l~lasma insulin. The shape of the early insulin response was not related to any meal characteristic. The potential biological and behavioral significance of the early insulin release is discussed. Insulin
Glycemia
Pre-absorptive reflexes
Cephalic phase
IN rats, an early preabsorptive release of insulin is observed within the first minutes following the intake of food [10,13]. This reflex response is biphasic: a first peak is of cephalic origin and appears within 3 minutes; a second peak follows between 3 and 6 minutes, and is likely elicited at the gastro-intestinal level. The amplitude of the insulin discharge appears to be dependent on the sensory qualities of the food presented: a more preferred food induces a larger cephalic insulin release than a low palatability food; furthermore, since this early insulin response disappears long before the end of the meal, it has been suggested that the amount of food eaten could be influenced by the taste-modulated insulin release [12]. In rats habituated to scheduled meals, the cephalic peak of the reflex insulin response can be conditioned to the mere appearance of the food cup [12] or to an arbitrary signal (time of day, or odor) that reliably predicts the imminence of food presentation [23,24]. A cholinergic mechanism has been shown to mediate the response as it is eliminated by atropine [24], and the determining role of the vagus has been ascertained by the disappearance of the cephalic response after vagotomy [10]. It has also been reported that the preabsorptive release of insulin facilitates and accelerates the post-ingestive glucose uptake in rats [ll]. In dogs, such a reflex insulin discharge has been observed within 10 minutes after an oral glucose load, while the plasma glucose level remained unchanged [5]. However satisfactory, this clear and coherent picture emerged from averaged insulinemia and glycemia curves which masked very large individual differences in the early insulin response to oral glucose. It has also been shown that dogs respond by
Food intake
Human subjects
a reflex insulin discharge to sham feeding of glucose, to tap water or cellulose presented orally [5, 6, 9]. In humans, a reflex insulin release of cephalic origin has been reported after visual and/or olfactory presentation of palatable foods in both lean and obese individuals [15, 18, 20], or even after hypnotic suggestion of food [8]. In these studies, the subjects were tested on a single, uniform stimulation condition. In spite of this, the insulin response often appeared to vary in amplitude and time course in different subjects. The most detailed study of this phenomenon in humans [20] included both lean and obese women and revealed that during visual and olfactive stimulation by food, the plasma insulin level could be highly elevated, somewhat elevated, or somewhat decreased relative to the level observed prior to food presentation. The proportions of the tests representing high, low, or negative cephalic insulin responses were not specified. In the same study, replication was attempted in obese subjects and the reflex insulin response exhibited poor within-subject reproducibility. In a sub-group of obese women, no change in the plasma insulin concentration were noted up to 20 rain after the visual presentation of a non food stimulus; this control was not reported for lean subjects. It is thus difficult to establish how frequently or reliably a given type of response is likely to appear in a normal population, or whether the amplitude of the response is a function of some subject's characteristic, or of the particular palatability conditions for a given individual. Since the particular type of food used in the experiment might have been differently palatable for different persons, it can be suggested that perhaps part of the variation in the cephalic response was due to differences in palatability. Be-
Copyright © 1983 Pergamon Press Ltd.--0031-9384/83/100515-07503.00
sir
!~t~[ ! iSl,l i : : :
sides, the relevance of such a reflex response for actual food consumption remains unknown. Obviously, studies of reflex, conditioned insulin release in humans present considerable technical difficulties since they involve blood sampling at short intervals, before, during and shortly after food presentation. Many factors associated with the procedure are likely to interfere with and modify the insulin level, stress being the most obvious. In all of the previous human studies, blood was collected at discrete intervals (1-10 min) shortly before, during, and shortly after the presentation of food stimuli and the reiteration of the blood sampling procedure may have acted as a factor of renewed tension or stress. Furthermore, the fact that the subjects were presented with food that they were not allowed to eat may have modified the cephalic response, as it alters the salivary response [25]. The occurrence of a preabsorptive, reflex insulin response has never been studied during actual consumption in human subjects. The reflex response which occurs in rats and humans on cephalic stimulation by food stimuli is likely to occur at the beginning of a meal. It might perhaps reflect palatability, and exert some influence on eating behavior. In the present experiment, the variations in insulinemia associated to food intake were examined in order to quantify such an eventual early insulin response in humans and to study its relationships with behavioral or metabolic events associated to the meal. Many studies in humans have examined the evolution of insulinemia and glycemia following different types of meals. Although these studies are highly informative about the differential absorption of certain types of foods, they use rather long blood sampling intervals (> 15 rain) that do not allow the observation of early reflex phenomena [2, 14, 16, 22, 26]. In the present experiment, seven human subjects were presented with many standard sandwich meals of different palatability. Their blood was collected continuously during and after the meal. Plasma insulin and glucose were assayed so as to obtain a fine-grain picture of their evolution from meal onset until 82 min later. The shape of the insulin curve in the first minutes of the meal, particularly before the glycemia curve started to reflect absorption, was followed closely. The amplitude of any early insulin variations and their relationships with palatability conditions, eating behavior, and the metabolic consequences of the meal were examined. Since subjects were tested repeatedly, the reproducibility of their early insulin responses was assessed. The experiment was conducted at Hopital Bichat (Paris) and Hopital du Kremlin-Bicetre. We would like to thank Dr. M. Apfelbaum and Dr. G. Barres for their cooperation. METHOD
Subjects The subjects were 4 women and 3 men, between 20 and 25 years of age. Their body weight was within -+ 10% of the ideal weight, according to the Metropolitan Life Insurance Co. Tables (1959). They were in good health at the time of the experiment. All were medical students who normally ate lunch at the hospital cafeteria. Therefore a sandwich lunch was not atypical for them. Before the beginning of the experiment, their alimentary habits and their attitudes towards food were assessed. The subjects had never been underweight or overweight in the past and had never had to diet. They exhibited no alimentary restraint; they habitually ate until complete satiation and never worried about their weight.
M~'ol (':,/;difion',
Three different meal conditions were presented ~wicc i~; each subject (total::6 meals per sutLjectL All meals were composed of standard, cocktail size sandwiche> These sandwiches, or food units, consisted of a 3 cm ~ piece ot commercial white bread, spread with a thin, uniform layer of a food of distinctive aspect and flavor. The nulritiomd value and texture of these food units, being determined largely hy the bread, were approximately equal lbr all typc~ of ~md wiches. (They represented approximately lt¢~ KJ :rod weighed about 7 grams.) In a mixed meal condition, five different spreads were presented: crab, anchovies, liver pate, pork pate and bu|ter. The other two meal conditions were high and low preference conditions, in which only the most preferred, ¢~ the lease preferred food was presented. The preference scale wa~ e~.tablished for each subject prior to the series of [ests, 013 the basis of a visual analogue test, as described ir~ a previous paper [ll.
Blood Sampling From an indwelling catheter (Vygon, 121 13, France) implanted in the subject's antecubital vein, blood was drawn in a continuous fashion by means of a peristaltic pump, at a flow rate of 0.8 ml per min, while a heparin solution (2000 tzl/ml saline) was delivered inside the catheter through a Butterfly needle (Abbott, 21-INT, Ireland) at a flow rate of 0.04 ml per min, so as to prevent clotting. Blood samples were collected minute per minute for the first 30 rain and then over 3 min intervals, up to a lotal of 84 min. They were centrifuged within a few minutes after drawing. Plasma was collected and stored at 24°C. Plasma insulin was measured in duplicate with the INSIK-1 radioimmunoassay kit (CEA, France). The interassay coefficient of variation was 12%. Plasma samples were assayed for glucose concentration with a YSI Glucose Analyser (YSI, 23A, Bioblock, France). The standard error with this method was + 1%.
Procedure The experimental sessions took place in quiet hospital rooms. The subjects were always tested alone. On test days, the subjects had their regular breakfast at the usual time, and were instructed to avoid any additional eating or unaccustomed physical activity before the test. The level of hunger was thus moderate at the beginning of the experimental sessions, and was expected to be representative of the normal deprivation conditions at lunch time. Each subject came to the experimental settings once 'a week, at 10:30 a.m. The subject was asked to sit in a comfortable armchair and the catheter was implanted immediately. It was kept patent through a saline perfusion (0.85%) which was continued for at least one hour before the beginning of blood drawing. This was done so that the effects of any stress due to the implantation of the catheter have disappeared by the time of blood drawing. No anaesthetics were used and none of the subjects showed or reported any mood perturbation at the time of catheterization. The subjects were then left to rest until the beginning of blood drawing, approximately 1 hour later. Immediately before blood drawing, the tip of the heparin delivering needle was inserted inside the catheter, and a constant heparin inflow to the catheter was started. The peristaltic pump was
R E F L E X I N S U L I N RESPONSE AND E A T I N G
517
then put into action to draw the heparinized blood. This represented no other puncture of the subject's arm, and the whole procedure was apparently tolerated without stress. Blood drawing was started at about 12:00 p.m. The blood sampled during the first 5-10 min was discarded. After the Experimenters felt sure that the blood and heparin flows were stable, the first 1 min blood sample was collected. The test food was prepared in an adjacent room a few minutes before being served, and was presented on a tray during the third minute of blood collection. The subjects were instructed to eat as much or as little as they wished. The duration and size of the meal were thus freely determined by the subjects. Prandial drinking was allowed ad lib. After the subject had stopped eating, the remaining food was removed. Meal size (number of sandwiches ingested) and meal duration were noted. Eating rate (number of sandwiches ingested per minute) was evaluated, since in a previous study [1] it was shown to be a very sensitive index of palatability. For one trial no food was presented, so as to measure the spontaneous variations of insulinemia and glycemia. The order of the 7 different trials was randomly determined. The Experimenters remained in the room during the whole session, including the meal.
/
linsuline
12108" i
i
i
i
I
I
i
.
1 2 3 4 5 6 "1 8 ; 10111' I~Z 1; 14 1!i 16 17 18 1~)2om|n FIG. i. Spontaneous variations of insulinemia and glycemia during 20 minutes under standard test conditions but without food presentation (means of seven tests_+SEM).
RESULTS Meal size and eating rate were compared between the three meal conditions. In the low preference condition 32_+4.1 sandwiches were ingested, and 37.5_+3.8 in the high preference condition. A two-way analysis of variance with repeated measurements revealed that these figures were not statistically different, F(1,6)=2.44, n.s. During mixed meals, the subjects ingested 44.9_+4.3 sandwiches. This is signifi-
Statistics
Analyses of variance and correlations were computed according to standard formulas [4]. Means_+SEM are presented in the text and figures.
120-
.glucose mg/m I
GLUCOSE mg/dl T
11010090-
80
NSULINE
-
-
",.~'/."
40-
30-
.
~
10-1 ~J
~,
lb
l~S
z,b
z's
3b
d.
~
~s
do
.g
~o
~
Yo
Ys
do M~N
FIG. 2. Variations in insulinemiaand glycemia during an 84 minute peri-prandial observation period (average of 42 meals +_SEM). Food was presented to the subjects during the 3rd minute of blood collection. The stars indicate the first sample significantly different from baseline (first 2 samples). While no significant change in the glucose level occurs before the 17th minute after the beginningof the meal, an early rise in plasma insulin becomes significant from the 4th minute after meal onset. After a pre-absorptive peak (Sth minute), plasma insulin decreases, but remains significantly elevated as compared to pre-meal values. After the 17th minute, insulinemiaand glycemia are closely parallel: they rise rapidly, reach a peal: level between 50 and 60 minutes after meal onset, after which they decrease slowly.
t~:l.I tSl.t: /: t ~i
518 .~
iNSULINE i~ U / m l
25
12 11
20
10 9 8
15-
7 6
10-
5 4 3
5-
2 1 0
~-.~..~_
__................................................................
o[-5O
-2'5
8
is
50
7'5
16o
I NSULINE
pU
-1
FIG. 4. Histogram presenting the distribution in amplitude of the cephalic insulin responses observed during 42 meals. The surface area of the insulinemia curve above or below baseline was integrated over the first 16 minutes after meal onset, before the appearance of any sign of absorption.
4 3 2 1
0 -1 -2
1 0
.....
1 -2
-1
o
t
2
3
4
5
6
7
s
9
lO
11 1 2 13 14 t 5
18
minutes
FIG. 3. Three types of preabsorptive insulin responses to the ingestion of food, as compared to baseline values. c antly more than in the low, F( 1,6) = 25.33, p < 0.01, and high, F(1,6)=6.5, p<0.05, preference conditions. Meal duration was 8.33_+3.5 minutes in the low preference condition and 10.7_+4.8 minutes in the hi_g_hpreference condition; these values are not significantly different, F(1,6)-=5.67, n.s. iVfixed meals lasted 11.8___3.8 minutes, which is significantly longer than low preference meals, F(1,6)=7.19, p<0.05, but not than high preference meals, F(1,6)=2.39, n.s. Eating rate (number of sandwiches ingested per minute) was identical in the three meal conditions: 3.9_+0.3 (low preference), 3.7_+0.2 (high preference) and 3.8_+0.2 (mixed meal): F(2,12)=0.31, n.s.
Figure 1 presents the spontaneous variations in insulinemia and glycemia at the hour of the test, without food presentation. Glycemia was extremely stable. While insulinemia was more variable, t tests for differences between correlated samples revealed that the two extreme points (3rd and 5th minutes) were not significantly different from each other, two-tailed t(6)=2.03, n.s. Figure 2 presents the evolution of insulinemia and glycemia associated to food intake (84 min observation period). These curves represent the average of the 42 meals. The glycemia graph is perfectly stable until the 17th minute; then a rapid increase occurs and glycemia remains elevated until the end of the observation period. On the basis of this observation, the continuous insulin and glucose curves were analyzed in two phases. The data gathered from the beginning to the 16th min of the meal were regarded as a first early phase, during which reflex fluctuations of cephalic or gastric origin might be observed. Data from the 17th min, up to the
82nd after meal onset, were regarded as a second phase, primarily describing absorption. T tests for difference between correlated samples revealed a significant elevation of the insulin level over baseline (average of the two minutes preceding food presentation), during the early phase. Insulinemia was elevated from the 4th min of the meal, two-tailed t(40) = - 3.13, p <0.01. On the 8th min after meal onset, insulinemia reached a peak value which was significantly elevated as compared to both baseline, two-tailed t(40)=5.64, p<0.001, and to the lowest point reached between this peak and the beginning of the absorptive phase, on the 12th rain after meal onset, twotailed t(40)=2.44, p <0.05. In turn, this last insulin value was significantly higher than baseline, two-tailed t(40)=3.62. p<0.001. An examination of individual curves revealed large differences in the insulin response. In the 42 experimental sessions, three main types of early insulin responses were observed: high and/or sustained elevation above baseline, low amplitude increase, and low amplitude decrease relative to baseline. The graphs presented in Fig. 3 illustrate these three early insulin patterns. A representative example is given of each type of response. Insulinemia and glycemia were analyzed in terms of the surface area above or below baseline (data observed prior to food presentation) for a given observation period (prior to or posterior to the rise in glycemia). The surface area of the early insulin response was computed for all 42 meals. The surface areas of insulinemia and glycemia beyond the 16th min of the meal were computed for all the tests in which blood sampling was successfully carried over for 84 min (N=34). Occlusion of the IV catheter prevented the collection of a complete series of data points in the other meals (N =8). Figure 4 shows a histogram of the amplitudes of the cephalic insulin responses, as integrated over 16 min. It appears that in only 5 out of the 42 meals (11.9%), the insulin curve was negative relative to baseline. In all other cases, the insulin curve was over baseline and the histogram reveals a continuous distribution of amplitudes, almost normal in shape. When no food was presented, the surface area of insulin variation around baseline, as integrated over 16 rain. was + 1 . 3 / z U (average of 7 tests). The profile of the cephalic insulin response was not linked in any simple fashion to subjects' characteristics or to meal
REFLEX INSULIN RESPONSE AND EATING composition: a given subject could exhibit largely different preabsorptive insulin responses, when presented repeatedly with the same food. During mixed meals, the surface area of the early insulin response was +31.5-+9.9/~U; in the high preference condition, it was +48-+9.1/zU; in the low preference condition, it was +50.7_+ 14.9/xU. A two way analysis of variance with repeated measurements indicated that these values were not significantly different: F(2,12)=0.81, n.s. The amplitude of the early insulin response was not correlated to meal size (number of food pieces ingested): r ( 4 0 ) = - 0 . 0 6 , n.s. ; to meal duration: r(40)=-0.188, n.s.; or to eating rate: r(40)=0.257, n.s. The positive early insulin response was usually perceptible for many minutes. Its time course was highly variable across different meals, even in the same subject. As shown in Fig. 3, one or two peaks could be observed before the sustained rise due to absorption. The evolutions of insulinemia and glycemia following the beginning o f absorption were different between test meals. Not only did the total surface area of insulin and glucose vary between meals, but the time course of the responses was variable. Generally, a sharp increase in plasma insulin and glucose was noted around the 16--18th min. In most cases, a high plateau was reached and the plasma concentrations started to decrease before the 84th min. In other cases, the increase in plasma insulin and glucose concentrations was slower to appear and/or no decrease was noted at the end of the observation period. The highest levels of insulinemia and glycemia were reached at unpredictable intervals after meal onset. The surface area of the early insulin response (first 16 min) was not found correlated to the surface area of the absorptive (17th-82nd min) insulinemia, r(32)=0.034, n.s., or glycemia, r(32)=0.118, n.s. The absorptive insulin surface area was not correlated to absorptive glycemia, r(32)=0.151, n.s.
Meal size was positively correlated to the amplitude of the integrated insulin response during the absorptive phase, r(32)=0.425, p <0.02, but not to the surface area of glycemia during the same period, r(32)=-0.088, n.s. DISCUSSION
An early phase of insulin release can be observed at the beginning of food intake in human subjects. In contrast with the stable insulinemia observed in the absence of food, an increase in plasma insulin, relative to baseline, was usually recorded within the first 16 min of food consumption, while the plasma glucose level remained unchanged. The total prandial and postprandial insulinemia curve therefore appeared to be clearly biphasic. In a first phase, which was contemporary with a stable glucose level, a variation in insulin level occurred a few minutes after the beginning of eating; insulinemia then receded towards baseline values, at about 10-12 min after meal onset. In a second phase, a sharp rise in insulinemia occurred concurrently with an increase in glycemia; this elevation was much more important than that of the first phase, and lasted over 60 min. This is consistent with prandial and post-ingestive insulinemia curves obtained in rats [10, I1, 21] and dogs [5, 6, 7]. In particular, the shape of the insulinemia curves presented here is very similar to that observed in dogs after an oral glucose load [5]. In these animals, one or two peaks of insulin could be observed before the rise in plasma glucose reflecting absorption: this also appeared in our human subjects as shown in the positive insulin responses presented in Fig. 3. Like our human sub-
519 jects, the dogs exhibited very irregular individual responses, even though they were simply presented with glucose. These observations suggest that the insulin variations which preceded the elevation in glycemia in our human subjects were preabsorptive and involved a reflex mechanism, which could be of cephalic and/or gastrointestinal origin. Of course, the early absorption of some nutrients before any variation was detectable in the glucose level cannot be completely ruled out. H o w e v e r the occurrence of similar insulin responses in the absence of ingestion or in sham feeding in various species [5, 8, 9, 11, 15, 18, 20] suggests the idea that the early insulin response observed here mainly reflects the action of reflex mechanisms in which cephalic stimulation could play a determinant role. Furthermore, the fact that the early increase in insulin is transient and recedes significantly before the clear concurrent elevation o f insulin and glucose is more easily understood if the early insulin response is considered the result of a preabsorptive reflex than that of a precocious absorption of nutrients by as yet unknown effectors: if the early insulin response reflected the beginning of absorption, it is difficult to see why insulinemia should exhibit one or two peaks, and then recede towards baseline values before a final sustained elevation. In spite of important differences in method (actual intake vs. visual stimulation; sandwich meal vs. varied hot meal), the early response reported here is extremely close to the reflex discharges observed in other human studies. Sjostrrm et al. [20] found comparable alterations in plasma insulin to occur during 5 min of actual visual and olfactory stimulation and to disappear on removal of the food stimulus. ParraCovarrubias et al. [15], working in obese adolescents, found a very high amplitude cephalic insulin response to outlast 15 min of visual and olfactory food presentation. The responses observed by Sjostrrm et al. [20] were almost immediate, appearing clearly within 1 or 2 min, while those reported by Parra-Covarrubias developed more slowly, and were still very modest after 5 min of stimulation. In the present study, the early insulin responses developed in a gradual manner over the first 8 min of food consumption and often extended beyond the duration of the meal. Differences in protocol could perhaps explain these different results. The deprivation time was longer in the Sjost r r m et al. [20] study and the protocol requested the subjects to skip a meal (breakfast) on test days. Although the obese subjects of Parra-Covarrubias et al. [15] were fasted overnight (10-12 hours), the experimental test was scheduled at normal breakfast time. Our subjects were only fasted for about 4 hours (between breakfast and lunch). Thus a more intense actual and/or perceived deprivation state might have led to steeper, more immediate cephalic responding in the Swedish study. This suggests that deprivation could exert a critical influence on the time-course of the cephalic responses, if not necessarily on their amplitude. The factors that command the amplitude or the shape of the reflex insulin response remain to be specified. The shape of the cephalic insulin curve does not appear to be determined in any direct fashion by stable subjects' characteristics, since a subject repeatedly presented with the same sandwich meal could exhibit largely different early insulin responses. In particular it does not appear that the subjects' " e x t e r n a l i t y " [17], or their relative sensitivity to food stimuli, could play a decisive role in determining the insulin secretion curve. On the other hand, immediate affective or cognitive alterations could partly determine the response. The amplitude of the cephalic insulin response was
520
BELI~ISI.I,I L7 Al
neither predicted by, nor predictive of any behavioral or physiological parameter, as measured in the present study. In particular, unlike what has been observed in rats by Louis-Sylvestre and Le Magnen [12], the amplitude of the integrated early insulin release did not vary coherently with palatability and was not found correlated to meal size. Two observations might help to explain these discordant results. First, it is clear that the palatability range in the two experiments was very different; in the study on rats, ingestion of a quinine-adulterated food was compared to that of a saccharin-edulcorated one; in humans, two unadulterated commercial foods were presented in sandwich form to the subjects. Contrary to what was observed in a previous work using a similar protocol [1], neither meal size nor eating rate, two sensitive indices of palatability, were significantly different between the two single flavor meals conditions in the present study, which indicates that probably the difference in palatability was extremely modest. This palatability range might not have been wide enough to produce important and systematic variations of the reflex insulin response. Secondly, it should be noted that in rats, the preabsorptive insulin response is of very short duration (less than 3 minutes) and has terminated before the end of the meal whereas in the present experiment the sandwich meals lasted for 10.3_+1 min on the average, and often terminated before the rise in glycemia signalling intestinal absorption. Even though most eating behavior was contemporary with the preabsorptive
physiological responses, no parameter of the meal wza~, shown to vary with the amplitude of the cephalic insulin response. Perhaps a longer meal, more varied in its com~ position, could ultimately be influenced by the shape of the early insulin response, although the mechanism of such a delayed effect of an early metabolic response on subsequent intake is difficult to imagine. As shown in Fig. 2, the early insulin response is extremely small in amplitude as compared to the post. absorptive secretion of insulin. In rats, the preabsorptive insulin response is proportionally greater, even in the case of a low palatability food [12]. The biological significance of such a weak response in humans, particularly its value as an anticipatory response to the absorption of nutrients, remains open to question, as it does not appear to reflect the level of palatability and predicts neither meal size, nor the future evolutions of insulinemia and glycemia. In humans as well as in animals, the prandial, preabsorptire insulin secretion response is probably only a part of the total neuroendocrine anticipatory response to the meal. For example, a cephalic phase of glucagon secretion [3~191 probably interacts with the insulin response to determine the subsequent physiological and behavioral responses. A more complete measurement of the total reflex response to food presentation might clarify its eventual impact on the ingestion and the absorption of a meal.
REFERENCES
1. Bellisle, F. and J. Le Magnen. The structure of meals in humans: Eating and drinking patterns in lean and obese subjects. Physiol Behav 27: 649-658, 1981. 2. Crapo, P. A., G. Reaven and J. Olefsky. Postprandial plasmaglucose and insulin responses to orally administered simple and complex carbohydrates. Diabetes 25: 741-747, 1976. 3. De Jong, A., J. H. Strubbe and A. B. Steffens. Hypothalamic influence on insulin and glucagon release in the rat. Am J Physiol 233: E380-E388, 1977. 4. Ferguson, G. A. Statistical Analysis in Psychology and Education. New York: McGraw Hill, 1971. 5. Fischer, U., H. Hommel, H. Fiedler and H. Bibergeil. Reflex mechanism on insulin secretion. Endocrinol Exp 8: 137-145, 1974. 6. Fischer, U., H. Hommel, M. Ziegler and E. Jutzi. The mechanism of insulin secretion after oral glucose administration. III Investigations on the mechanism of a reflectoric insulin mobilization after oral stimulation. Diabetologia 8: 385-390, 1972. 7. Fischer, U., H. Hommei, M. Ziegler and R. Michael. The mechanism of insulin secretion after oral glucose administration. I Multiphasic course of insulin mobilization after oral administration of glucose in conscious dogs. Differences to the behavior after intravenous administration. Diabetologia 8: 104-110, 1972. 8. Goldfine, I. D., C. Abraira, D. Gruenwald and M. S. Goldstein. Plasma insulin levels during imaginary food ingestion under hypnosis. Proc Soc Exp Biol Med 133: 274-276, 1970. 9. Hommel, H., U. Fischer, K. Retzlaff and H. Knofler. The mechanism of insulin secretion after oral glucose administration. II Reflex insulin secretion in conscious dogs bearing fistulas of the digestive tract by sham-feeding of glucose or tap water. Diabetologia 8:111-116, 1972. 10. Louis-Sylvestre, J. Preabsorptive insulin release and hypoglycemia in rats. A m J Physiol 230: 56--60, 1976. 11. Louis-Sylvestre, J. Relationship between two stages of prandial insulin release in rats. Am J Physiol 235: E103-EI 11, 1978.
12. Louis-Sylvestre, J. and J. Le Magnen. Palatabability and preabsorptive insulin release. Neurosci Biobehav Rev 4: Suppl I, 43-46, 1980. 13. Louis-Sylvestre, J. and J. Le Magnen. Phase cephalique de secr6tion d'insuline et vari6te des aliments au cours du repas chez le rat. Reprod Nutr Ddv, in press. 14. O'Dea, K., P. J. Nestel and L. Antonoff. Physical factors influencing postprandial glucose and insulin responses to starch. Am J Clin Nutr 33: 760-765, 1980. 15. Parra-Covarrubias, A., I. Rivera-Rodriguez and A. AlmarazUgalde. Cephalic phase of insulin secretion in obese adolescents. Diabetes 20: 800-802, 1971. 16. Potter, J. G., K. P. Coffman, R. L. Reid, J. M. Krall and M. J. Albrink. Effect of test meals of varying dietary fiber content on plasma insulin and glucose response. Am J Clin Nutr 34: 328334, 1981. 17. Rodin, J. The role of perception of internal and external signals on the regulation of feeding in overweight and non-obese individuals. In: Appetite and Food Intake, Lift, Sciences Report 2, edited by T. Silverstone. Berlin: Dahlem Konferenzen, 1976, pp. 265-283. 18. Sahakian, B. J., M. E. Lean, T. W. Robbins and W. P. T. James. Salivation and insulin secretion in response to food in non-obese men and women. Appetite 2: 209-216, 1981. 19. Samoles, E., J. Tyler, V. P. Rege and V. Marks. The Possible Role of Glucagon (Pancreatic and Extrapancreatic) in Insulin Secretion. Stockholm: VIth Congress International Diabetes Federation, 1967, pp. 446-454. 20. Sjostr6m, L., G. Garellick, M. Krotkiewski and A. Luyckx. Peripheral insulin in response to the sight and smell of food. Metabolism 29: 901-909, 1980. 21. Steffens, A. B. Plasma insulin content in relation to blood glucose level and meal pattern in the normal and hypothalamic hyperphagic rat. Physiol Behav 5: 147-151, 1970.
REFLEX
INSULIN RESPONSE
AND EATING
22. Vaaler, S., K. F. Hanssen and O. Aagenaes. Plasma glucose and insulin responses to orally administered carbohydrate-rich foodstuffs. Nutr Metab 24: 168-175, 1980. 23. Weiley, J. H. and G. A. Leveille. Significance of insulin in the metabolic adaptation of rats to meal ingestion. J Nutr 100: 1073-1080, 1970. 24. Woods, S. C., J. R. Vasselli, E. Kaestner, G. A. Szakmary, P. Milburn and M. V. Vitiello. Conditioned insulin secretion and meal feeding in rats. J Comp Physiol Psycho191: 128--133, 1977.
521
25. Wooley, S. C. and O. W. Wolley. Salivation to the sight and thought of food: a new measure of appetite. Psychosom Med 35: 136--142, 1973. 26. Zaccaria, M., E. De Palo, E. Zago, N. Sicolo, G. Erie and G. Federspil. Metabolic and endocrine responses to a standard mixed meal. A physiologic study. Acta Diabetol Lat 16: 45-53, 1979.