Effect of test meals of varying dietary fibercontent on plasma insulin and glucoseresponse1
Joan G. Potter, M.D., Kathryn P. Coffman, B.A., Robert L. Reid, Ph.D.,
John M. Krall, Ph.D., and Margaret J. Albrink, M.D.
ABSTRACT To assess the effect of dietary fiber on glucose tolerance four different meals of
varying fiber content but identical protein fat and carbohydrate content were fed to eight healthy
men aged 22 to 45. Each meal provided 75 g of carbohydrate as liquid glucose formula, as brown
rice, pinto beans, or All Bran. The mean plasma glucose and insulin responses were highest
following the formula, and least for All Bran and pinto beans. Rice produced nearly as great a rise
in insulin and glucose as did the formula. The rank of each meal by content of neutral detergent
fiber was nearly the inverse of the rank by magnitude of the insulin response evoked, fiber content
being greatest in All Bran (18 g) and pinto beans (16.2 g), low in rice (2.8 g) and absent from the
formula. It was concluded that dietary fiber dampened the insulin response to a high carbohydrate
meal. Am. J. Clin. Nutr. 34: 328-334, 1981.
KEY WORDS Dietary fiber, glucose tolerance, plasma insulin, triglyerides. test meals, plasma
lipids
Hyperinsulinemia is associated with hyper-triglyceridemia (1), obesity (2), and athero-sclerotic heart disease (3). Because the hyper-insulinemia might play a role in the etiologyof these conditions factors regulating plasmainsulin concentrations are of interest. Al-though between-individual variation in theinsulin response to a standard glucose chal-lenge reflects endogenous factors governinginsulin release, different sources of dietarycarbohydrate may evoke widely differing in-sulin response in the same individual (4-6).
Certain properties of food thus influence theinsulin response to a meal. We recently re-ported that a high carbohydrate diet in whichthe carbohydrate was given as foods rich infiber did not cause carbohydrate-induced 11-pemia while carbohydrate given in a liquid
formula caused the expected increase in tri-glycerides (4). The insulin response to glucosegiven as the liquid formula was much greaterthan to a test meal containing an equalamount of carbohydrate as brown rice andcooked dried beans. It was postulated that
the high fiber content of the meals, by blunt-ing the insulin response to the meals, was
328
responsible for the lack of rise of fastingtriglycerides after one week on the diet.
The present study was undertaken to eval-uate the response of plasma glucose and in-sulin to the chief high fiber foods used in thatstudy: wheat bran cereal, pinto beans, and
brown rice, compared to a liquid formulameal. The results showed marked variationin fiber content and negative correlation be-tween dietary fiber content and both plasmaglucose and insulin response to the meal.
Methods
The experiment was approved by the West Virginia
University Committee on Protection of Human Subjects.
Informed signed consent was obtained from each subject.
‘From the Department of Medicine and the Depart-
ment of Community Medicine and the College of Agri-
culture and Forestry. West Virginia University. Morgan-
town, West Virginia 26506.
2 Supported in part by Institutional Grants 2-210-1615
(75) and 5 K R01 AM09252 and by Research Career
Award 5K6HL00486. NHLBI (M.J.A.).
Address reprint requests to: Margaret J. Albrink.
M.D., Department of Medicine. West Virginia Univer-sity Medical Center. Morgantown. West Virginia 26506.
The subjects were 8 healthy men aged 22 to 45. Nonewas obese. Mean relative weight was 100. range 9 1 to
I 10. Each received four test meals, about a week apart.in latin square design. Thus the order in which the meals
were consumed was evenly distributed amongst the sub-
jects.
Meals
Composition of the four test meals is shown in Table
1. The amount of food in the test meal was designed to
provide 75 g of carbohydrate (7). Casein (Casec. Mead
Johnson) or corn oil (Mazola) were added as necessary
to equalize the protein and fat content of all meals. Theingredients of the formula meal were mixed in a blender.
Each subject was required to ingest his meal continuously
over 20 mm and to add enough water to each meal so
that the total volume of water. 400 ml including foodwater, was the same in each meal. The meals were
prepared in batch and frozen until needed. The beans
and rice were cooked by steaming. An aliquot of each
meal was saved for fiber analysis.
The subjects consumed their usual diets before and
between the test meals. They pursued their usual semi-
sedentary lives during the tolerance tests.
Laboratory procedures
Blood was drawn fasting. at 30. 60. 120, and 180 mm
after the beginning of ingestion of the test meals. Each
sample was analyzed for insulin (8) and glucose (Beck-
man glucose analyzer). Cholesterol (9) and triglycerides
(10) were measured on the fasting and 3-h samples.
Aliquots of uncooked pinto beans, brown rice, and
All Bran were analyzed for dietary fiber by a modifica-
tion of the method of Van Soest (II. 12). Total cell wall
fiber was analyzed directly as neutral detergent fiber
(NDF) which includes lignin. cellulose, and hemicellu-
lose. Acid detergent fiber (lignin + cellulose) and lignin
were also analyzed directly while cellulose and hemicel-lulose were calculated by difference.
Statistical analysis was carried out by analysis of
variance, the order of the meal and type of meal being
included as factors which were tested. Tukey’s multiple
comparison procedure was used to locate the significant
differences for each diet for each time interval after
ingestion of the test meal (13). In certain instances paired
test was used.
Results
There was no effect of order of meal but alarge effect of type of meal. The plasmaglucose and insulin responses to the test mealsare shown in Figure 1. The chief differenceswere in the ‘/2 and 1 h values, where theformula produced the greatest rise in plasma
glucose, rice being next, All Bran and pintobeans causing the least rise. The significantdifferences using Tukey’s test are shown in
the legend. Of interest also was the three hourglucose value which for the formula diet was
less than for the other three meals and the
difference was significant when tested bypaired I test. A specific t test was calculatedat 3 h because the standard errors were con-siderably smaller at that time than those at 0,�/2, 1, and 2 h.
The mean insulin response to the formulawas nearly twice that to the beans and the All
Bran. The rank of insulin response to the fourmeals was the same as that of plasma glucose,the formula causing the greatest rise in insu-lin, rice causing a slightly lesser response, AllBran and beans causing the least response.The peak insulin response to pinto beans waslater than for the other three meals.
The difference in insulin response wasmuch greater than the difference in glucoseresponse, a nearly 2-fold greater peak insulinoccurring after the formula meal than afterthe pinto beans and All Bran meals. The peakglucose response to formula was only 27%greater than the response to All Bran andpinto beans.
The mean plasma triglyceride concentra-tion at 3 h was lower than the fasting concen-tration after all four meals, significant for two
of the four meals. The decrease was greatestafter the formula and least after the pintobeans (Fig. 2). The triglyceride decrease wasroughly correlated with the insulin and glu-
cose rise stimulated by the meal (Fig. 1). The‘/2 h insulin bore a low grade correlation withthe percent decrease in triglyceride (r = 0.331,
p < 0.05). The mean and SD of fasting cho-
lesterol before and after all four tests togetherwas 184 ± 37 and 186 ± 35 mg/dl and it didnot change significantly after any meal.
The type and amount of fiber varied
greatly among the three types of food. TablesI and 2 show the amount and composition ofthe meals, each designed to supply 75 g ofcarbohydrate. All Bran and pinto beans,which were similar in evoking the least re-sponse of glucose and insulin, also had thegreatest amount of NDF (Table 2). Brownrice by comparison had only a small amountof NDF. All Bran and pinto beans were
similar in lignin content but All Bran hadmore hemicellulose than cellulose while pintobeans had more cellulose than hemicellulose.
The ranking by insulin response was the in-verse to ranking by NDF, cellulose, or lignin.The ranking by book value for crude fiber (7)
FIG. 1. Response of plasma glucose and insulin to various meals containing 75 g of carbohydrate. means and
SEM. Significant differences by Tukey’s test as follows:
Probability values
Formula versus bran
Formula versus beans
Rice versus bran
Rice versus beans
40
2 3
Glucose Insulin
‘/.h Ih 3h ‘/ih Ih 2h
p < 0.01 <0.01 <0.01* <0.01 <0.01 <0.10
p < 0.01 <0.05 <0.01 * <0.01 <0.10
<0.05
p<O.IO <0.01
330 POTTER ET AL.
* Not significant by Tukey’s test but significant by paired t test.
was similar to the ranking by content ofNDF.
Discussion
The four types of meals, containing thesame amount of protein, fat, and carbohy-drate, but varying in fiber content, evokedvery different responses of plasma glucoseand insulin. The liquid formula caused thegreatest response of both insulin and glucose,
and All Bran and pinto beans caused the leastresponse. Rice caused almost as great a riseas the liquid formula. The differences weregreater for insulin than for glucose. Factors
that might influence insulin response such asthe time required for ingestion of the meal,amount of fluid, and amount of protein andfat were strictly controlled and were identicalin all meals. The monosaccharide, either asfed or as the product of digestion, was glucosein all four meals. The peculiar property of
fructose of causing little increase in plasmaglucose or insulin was thus avoided (14).
The fiber content, as NDF, of the mealsvaried inversely with the insulin and glucose
response, being absent from the liquid for-mula and highest in the pinto beans and All
Bran.We earlier reported that refined wheat
starch as bread evoked a slightly lower insulin
and glucose response than an equivalentamount of ingested glucose (14). Some of thedifferences in the insulin effect of the mealsin the present study could be accounted forby the difference between glucose in solutionand cooked starch. However, the large differ-ence between the rice, which evoked a highinsulin response, and All Bran and pintobeans, which evoked a low response, must bedue to other factors.
Another possibility is that the solid meals
contained less than the intended 75 g ofcarbohydrate. The amount of carbohydratewas calculated from the book values (7) inwhich carbohydrate is calculated by subtract-
ing water, protein, fat, and ash from totalweight. The NDF would thus be included inthe designation “carbohydrate” and would
displace nearly 20 g of the calculated 75 g ofstarch in the All Bran and bean meals, butonly 2 to 3 g in the rice meal. While most ofthe fiber components would be converted toa digestible form by the action of bacteria inthe colon and would thus provide calories(15) they would not provide a plasma glucose
or insulin raising stimulus. Peterson andReaven (16) showed double or more the in-sulin response to 80 grams compared to 40 gof glucose. The 2-fold difference between thetwo glucose doses was much greater than thepostulated difference in starch between pintobeans and formula created by displacementby fiber in the present study. Thompson eta!. (17) found no difference in integratedglucose or insulin response whether meals
contained 65 or 45% of calories from carbo-hydrate as corn syrup. However, a small
amount of the difference in response to ourvarious meals could be due to differences inamounts of starch.
Protein also stimulates insulin release (18).Care was taken to equalize protein content ofall meals by adding casein where needed(Table 1). The casein could have been more
insulinogenic than the vegetable protein itreplaced. The fact that the insulin responseto the bran and bean meals was similar (Fig.1) while the casein content was very dissimilar(Table 1) makes it unlikely that casein was
responsible for the difference.Degree of cooking and physical state could
account for some of the differences in theeffect of the meals. Cooked starch meals raiseplasma glucose and insulin while raw soluble
starch does not (6) although it supports ca-loric balance (19). A meal of whole applescaused a lesser increase in plasma glucoseand insulin than an equivalent amount ofapple puree although both forms were isofi-brous (20). Apple puree caused less rise than
apple juice suggesting that sugar in solutionis particularly insulinogenic (20). The foodsin the present study were cooked to a palat-able state, but by their nature differed intexture. The effect of degree and type ofcooking on insulin response to meals remainslargely unexplored. While the difference be-tween liquids and solid food could have ac-counted for some of the difference betweenour formula and solid foods it could notaccount for the differences between rice andthe other solid foods.
Of all the variables discussed, dietary fiberseems most likely to account for the greatlydifferent responses, the rank of insulin andglucose response being the inverse of the rankwith regard to NDF content. All Bran andpinto beans were nearly indistinguishablewith respect to NDF and the plasma glucose
and insulin response evoked. The ranking bymagnitude of insulin response correlated bestwith ranking by cellulose and/or lignin con-tent.
The range in insulin response between themean peak for the rice meal and that for the
bean and bran meals, about 2-fold, was muchsmaller than the 10-fold difference betweenthe NDF content of rice (2.8 g) and All Bran
(20 g). If fiber is responsible for any of thedifference, the amount of fiber in the beans
and All Bran may be far in excess of thatneeded to reduce insulin output.
Crapo et a!. (5) also reported differing glu-
cose and insulin response to various mealscontaining equal equivalents of glucose andin their study as in ours insulin responsediffered more than did glucose response. In-
sulin response to rice and corn was lowerthan to glucose solution while potatoesevoked almost as great an insulin response asdid glucose, and white bread was intermedi-ate. A discrepancy between their study andthe present study is that insulin response tobrown rice in our study was greater than towhite rice in their study. Differences in fiber
between their brand (Uncle Ben’s ConvertedRice) and our brand (River Brand BrownRice) or in cooking could account for the
discrepancy. Rank of insulin response in theirstudy was nearly the inverse of rank of pub-lished values of cell wall fiber determined byVan Soest (21) with the exception of ricewhich was not included in that list. If rice is
assumed to be low in fiber, as our Table 2,then the low insulin response to rice in theirstudy would not be explained by the low fibercontent. Although few other studies have cor-related insulin response with fiber content ofmeals, Jenkins et al. (22) reported that addi-
tion of guar and pectin, only variably de-tected by the methods of the present study,blunted the insulin response to glucose chal-
lenge (22). By creating less demand for insu-un, or by other unknown mechanisms, highfiber diets have been beneficial in the treat-ment of diabetes (23, 24).
The mechanisms by which high fiber dietslower insulin response is not known. Theimportance of viscosity in causing delayedgastric emptying time and slowing of carbo-hydrate absorption has been suggested (25),
although viscosity of bran could not be mea-sured. The fact that the difference in insulinresponse was much greater than the differ-ence in glucose response could be an indica-tion of enhanced insulin sensitivity followingthe high fiber meal, through unknown mech-
A role for excess insulin response in thegenesis of hypertriglyceridemia has been pos-tulated (1). The rise in fasting plasma triglyc-erides in normal persons after several days ofvery high carbohydrate diet (carbohydrate-
induced lipemia) has also been attributed tothe excess insulin secretion stimulated bysuch diets (I). The role of insulin isstrengthened by our earlier finding that whenthe high carbohydrate diet was administeredas foods high in fiber, a representative meal
stimulated only a slight rise in insulin. and acarbohydrate-induced rise in fasting triglyc-erides, while it occurred following a low fiberdiet, did not take place (4).
When hypertriglyceridemia is induced by
a high carbohydrate diet the hypertriglyceri-demia is most marked in the fasting state,decreasing after the first high carbohydrate
diet of the day, and remaining low through-out the day (26). The magnitude of acutedecrease in triglycerides with feeding wasfound by Barter et al. (26) to be related to themagnitude of insulin response to the meal.
The authors postulated that fasting hypertri-glyceridemia induced by prior high carbo-hydrate diet represented impaired triglycer-
ide removal at an insulin sensitive site, prob-ably adipose tissue. In the subjects of thepresent study a similar reduction of triglyc-
eride concentration below fasting levels oc-curred 3 h after ingestion of the meals. Thereduction was greatest after the glucose for-mula meal, and least after the bean meal. Thereduction of triglyceride levels with carbo-hydrate feeding thus occurs not only in sub-jects with carbohydrate-induced lipemia butalso in persons with normal triglycerides whohave previously been consuming ordinarydiets. The acute triglyceride-lowering effectof glucose or insulin administration and itsprobable dependence on insulin-stimulated
triglyceride removal has been reviewed byMacdonald (27). The greater postprandial
decrease in triglycerides after the high car-bohydrate low fiber meals, about 30%, thanafter the high fiber meal, about 10%, mightbe expected to result in lower day-long tri-
glyceride levels with low than high fiber dietswere it not for the fact that after a few days
of high carbohydrate low fiber diet a rise in
fasting triglycerides of at least 30% would beexpected to occur (4). We found that post-prandial triglycerides as well as fasting tri-glycerides were higher during a low fiber thanhigh fiber diet (4), although in that study thegreater fat content (15% of calories) than in
the present study (8% of calories) obliteratedthe postprandial fall in triglycerides.
The acute decline in triglyceride concentra-tion after glucose or insulin administrationmay represent a normal shift from the fateconomy of fasting, in which adipose tissueis relatively inactive in removing circulating
triglycerides, to the carbohydrate economy ofthe fed state with its increased insulin-stimu-lated triglyceride removal by adipose tissue.The fact that the magnitude of triglyceride
reduction was related to the magnitude ofinsulin response, in our study as in that of
Barter et al. (26), offers an explanation forthe trend for less triglyceride reduction afterthe high fiber meals than after the formulameal. Dietary fiber may thus dampen the
shift to carbohydrate metabolism with feed-ing, a possibility strengthened by the occur-rence of reactive hypoglycemia after the fi-
ber-free formula meal (Fig. 1).To the extent that dietary fiber dampens
insulin response, a high fiber-diet would re-sult in lower ambient insulin concentrationsduring the day, a more gradual and perhapsdiminished shift from the fat economy offasting to the carbohydrate economy of thefed state, a more even 24-h metabolic state,a lack of reactive hypoglycemia, and a stabletriglyceride concentration during the day. Ifthe suggestive evidence of the present studythat insulin response could be predicted fromfiber content is borne out, diets could betterbe designed for the treatment and prevention
of diseases in which excessive insulin re-sponse might have an etiologic role, such asobesity (2), hypertriglyceridemia (1), and ath-erosclerosis (3). a
The skillful laboratory assistance of Ruth Deal is
gratefully acknowledged.
References
I. Olefsky JM, Farquhar JW. Reaven GM. Reap-
praisal of the role of insulin in hypertriglyceridemia.
Ami Med 1974:57:551-60.
2. Assimacopoulos-Jeannet F. Jeanrenaud B. The hor-
monal and metabolic basis of experimental obesity.
Clin Endocrinol Metab 1976:5:337-65.
3. Stout RW. Diabetes and atherosclerosis-the role of
insulin. Diabetologia 1978:16:141-50.
4. Albrink Mi, Newman T, Davidson PC. Effect of
high-and low-fiber diets on plasma lipids and insu-
lin. Am J Clin Nutr 1979:32:1486-91.
5. Crapo PA, Reaven 0. Olefsky J. Postprandial
plasma-glucose and insulin response to different
complex carbohydrates. Diabetes 1977:26:1178-83.
6. Crapo PA, Reaven 0, Olefsky J. Plasma glucose and
insulin responses to orally administered simple and
complex carbohydrates. Diabetes 1976:25:741-47.
7. Composition of Foods. Agriculture Handbook no. 8.
Superintendent of Documents. Washington. D.C.:United States Government Printing Office, 1963.
8. Morgan OR, Lazarow A. Immunoassay of insulin:
two antibody system. Diabetes 1963:12:115-126.
9. Abell LL, Levy BB, Brody BB. Kendall FE. A
simplified method for the estimation of total choles-
terol in serum and demonstration of its specificity. J
Biol Chem 1952:195:357-66.
10. Lofland HB Jr. A semiautomated procedure for the
determination of triglycerides in serum. Anal Bio-
chem 1964:9:393-400.
II. Goering HK, Van Soest PJ. Forage fiber analyses.
Agricultural Handbook no. 379. Washington. D.C.:
ARA, USDA, 1970.
12. Fonnesbeck PV. Partitioning of the nutrients of for-
age for horses. J Anim Sci 1969:28:624-33.
13. Winer Bi. Statistical principals in experimental de-
sign. 2nd ed. New York: McGraw-Hill Book Co..
1971:196-201.
14. Swan DC, Davidson PC, Albrink Mi. Effects of
simple and complex carbohydrates on plasma non-
esterified fatty acids, plasma sugar and plasma in-
sulin during oral carbohydrate tolerance tests. Lan-
cet 1966:1:60-3.
15. Bond JH, Levitt MD. Fate of soluble carbohydrate
in the colon of rats and man. I Clin Invest 1976:57:
1158-64.
16. Peterson DT, Reaven CM. Evidence that glucoseload is an important determinant of plasma insulin
response in normal subjects. Diabetes 1971:20:729-
33.
17. Thompson RG. Hayford IT. Danney MM. Glucose
and insulin response to diet. Effect of variations insource and amount of carbohydrate. Diabetes
