David J.A. Jenkins, MD, PhD Thomas M.S. Wolever, MD, Starchy Foods and PhD Glycemic Index Alexandra L. Jenkins, RD

Different starchy foods produce different glycemic the glycemic impact of the diet could be kept constant responses when fed individually, and there is some regardless of the variety of foods used (3,4). evidence that this also applies in the context of the Later studies by Crapo and colleagues (5-8) focused mixed meal. A major reason appears to relate to the on the differences between starchy foods of similar mac- rate at which the foods are digested and the factors ronutrient composition. Differences in both and influencing this. A similar ranking in terms of glycemic response to specific foods is seen independent of the responses were observed, and it was postulated carbohydrate tolerance status of the groups tested. that possible differences in rates of digestion of the foods Potentially clinically useful starchy foods producing were responsible. These differences in rates of digestion relatively flat glycemic responses have been identified. of starchy foods were subsequently confirmed (Fig. 1) Many of these are considered ethnic or traditional and and related to the glycemic responses observed in both include ; ; such as , parboiled normal and diabetic individuals (9,10). , and bulgur (cracked wheat); and whole- From the beginning of the 1980s, many tests of single breads such as . Specific incorporation foods (11-28) and mixed meals (24,29-35) have been of these foods into diets has been associated with undertaken in both normal and diabetic subjects. How- reductions in low-density lipoprotein cholesterol and triglyceride levels in hyperlipidemia and with improved ever, because of a lack of standardization of methods blood glucose control in insulin-dependent diabetic of data presentation, the results of different studies were patients. To facilitate identification of such foods, it has not always directly comparable. In 1981, the concept been suggested that the glycemic response should be of the glycemic index (Gl) was proposed as a method indexed to a standard (e.g., white bread) to allow of assessing and classifying the glycemic response to comparisons to be made between the glycemic index of carbohydrate foods (11). It was hoped that this would foods tested in different groups of subjects. The scope allow foods to be compared more readily. It would also of application of this principle is subject to further allow the experience of different investigators to be investigation. It may be used to expand the range of pooled by indexing the foods tested to a common stan- possibly useful starchy foods for trial in the diets of dard. Initially, glucose was used, but this proved to be diabetic patients. Care 11:149-59, 1988 less acceptable for routine use than white bread of known composition. The Gl was therefore defined as

incremental blood glucose area after food x 100 ifferent carbohydrate foods produce different corresponding area after equicarbohydrate portion of white bread glycemic responses despite an apparent lack of difference in macronutrient composition (1,2). The classification of carbohydrate foods D From the Department of Nutritional Sciences, Faculty of Medicine, and the was first put on a systematic basis by Otto and col- Division of Endocrinology and Metabolism, St. Michael's Hospital, University leagues (3,4), who, after testing foods, allowed carbo- of Toronto, Toronto, Ontario, Canada. Address correspondence and reprint requests to David J. A. Jenkins, Depart- hydrate incorporation into the diabetic diet in propor- ment of Nutritional Sciences, Faculty of Medicine, University of Toronto, To- tion to the glycemic response they produced. In this way ronto, Ontario M5S 1A8, Canada.

DIABETES CARE, VOL. 11, NO. 2, FEBRUARY 1988 149 GLYCEMIC INDEX

^ 7 — ^^* Wholemeal Bread

•^ E / Miolcn- E E 5 y^ ^^^l-Tlte Spaghetti Rice y' /^^^^Nuict W n 4 / ^ ^S^, Wliolen eal Spaghetti p.c uO ^ ^^ y^^/yr ^^^^^^ Porrid i3 fat Peas 2-5 2

FIG. 1. Increase in concentration over 5 lysat e bohy d h of products of digestion, mea- o.2 0 sured as glucose after acid hydrolysis, subsequent to incubation of 2 g available .E 0 12 3 4 carbohydrate portions of foods with pooled human saliva and pancreatic Time (hours) juice.

By applying this approach to data from different groups made (38). However, when considering the relative gly- of subjects and different centers (12), it has been pos- cemic effects of different foods, i.e., the glycemic index, sible to begin to classify a substantial number of foods there is in fact some evidence for agreement among in terms of their glycemic responses (Table 1). different groups (Table 1). Foods that have been shown to have low glycemic Early studies with four starchy foods (bread, potato, responses include whole-grain (as opposed to whole rice, and corn) demonstrated the same order of ranking meal) cereals (15), pasta (17,36), and legumes (13,21,37). of the glycemic and insulin responses when these foods It was suggested that inclusion of such foods in the diets were tested in nondiabetic compared with diabetic vol- of patients with diabetes might aid dietary management unteers (6,8). Since then, several studies have shown by improving diabetes control. similarities in the ranking of responses to a wide range of foods tested in nondiabetic, NIDDM (13), and insu- lin-dependent diabetic (IDDM; 15-1 7) subjects. On the other hand, many studies do not agree (Table 1). OBJECTIONS TO GLYCEMIC INDEX More recently, it has been maintained that consider- ation of average glycemic responses is inadequate be- Objections to the Gl concept were raised early (38) and cause they may conceal large differences in response in have not been resolved (34,39). These objections have different individuals (39). This objection would be of resulted in a statement from the recent NIH consensus major clinical importance if the variability in response conference on diet and exercise in non-insulin-depen- between patients was such that certain individuals con- dent diabetes (NIDDM) that recommended against the sistently failed to show the expected differences in gly- use of Gl in the dietary management of diabetes (40). cemic responses between foods. The prescription of a The concern revolves around 3 major issues: 7) large diet containing foods of lower Gl would certainly not individual variation in responses, 2) lack of agreement result in lower postprandial blood glucose responses among different centers, and 3) lack of difference be- throughout the day. Unless these individuals could be tween mixed meals. In addition, it has been pointed out readily identified, the clinical application of Gl data that there are no studies showing long-term benefits of would indeed be limited and inappropriate if the num- low-GI foods (38,40). For these reasons it has been ber of patients who failed to show a consistent response maintained that the Gl has no clinical utility (34,38- was large. In view of the substantial coefficient of vari- 40). ation often seen in the GI to single foods, this negative Individual variation in glycemic responses. There are outcome is a real possibility. large differences among individuals with respect to the We have therefore examined the individual data that absolute level of blood glucose achieved after meals. formed the basis for recently published papers. In these Factors that have been suggested to influence this in- studies several low-GI foods were taken by different di- clude the presence and type of diabetes (38,41,42), age, abetic patients (Table 2; 15,16). Such a range of foods sex, body weight, and race (40). It has therefore been might be exchanged for foods of higher Gl in the diets stated that glycemic responses to foods should be tested of diabetic patients. We therefore considered it clinically in the specific group for which recommendations are relevant to determine whether the overall response to

150 DIABETES CARE, VOL. 11, NO. 2, FEBRUARY 1988 D.J.A. JENKINS, T.M.S. WOLEVER, AND A.L JENKINS

TABLE 1 Mean glycemic index (Gl) values of foods adjusted proportionately so that Gl of white bread = 100

Food Gl values* Subjectst Mean Gi

Breads Crispbread 90", 100 C,A 95 Whole meal 89 G 89 , i.e., pumpernickel 58", 78a C,G 68 Wheat White 100 (defined) A-K,M 100 Whole meal 93", 96, 100, 104, 106 C,G,B,A,J 100 ± 2 Pasta Macaroni White, boiled 5 min 64" I 64 Spaghetti Brown, boiled 15 min 61" A 61 White, boiled 15 min 46", 59", 68", 72C I,B,C,A 61 ± 6" White, boiled 5 min 45" I 45 Protein enriched 38b I 38 Star pasta White, boiled 5 min 54" I 54 Cereal grains Barley (pearled) 31C F 31 Buckwheat 74 A 74 Bulgur 65C G 65 103 A 103 Rice Brown 65", 96 N,A 81 Instant, boiled 1 min 65" H 65 Instant, boiled 6 min 121" L 121 Polished, boiled 5 min 58C H 58 Polished, boiled 15 min 68-, 70", 73", 78a, 83", 104 M,L,C,B,H,A 79 ± 5" Parboiled, boiled 5 min 54C H 54 Parboiled, boiled 25 min 58", 66", 72C, 78" D,H,E,R 65 ± 4J Rye kernels 47C G 47 Sweet corn 66C, 67" 85", 86, 87, 90 E,L,R,A,J,D 80 ± 4" Wheat kernels 63C G 63 Breakfast cereals All Bran 71", 74a, 76" B,A,N 74 ± 1b Cornflakes 107", 116, 121" L,A,B 115 ± 4 Muesli 96 A 96 Porridge oats 71a, 88", 93, 96 A,C,K,B 87 ± 6 Puffed rice 132" L 132 Shredded wheat 97 A 97 Weetabix 109 A 109 Cookies Digestive 77, 86 B,A 82 Oatmeal 78 A 78 Rich tea 80 A 80 Plain crackers (water biscuits) 91 A 91 Root Potato Instant 116, 119" A,L 118 Mashed 100 J 100 New, boiled 67", 75, 78", 101 C,B,L,A 80 ± 7 Russett, baked 112, 134~, 137a E,R,D 128 ± 8 Sweet 70a A 70 74 A 74 Legumes Baked beans (canned) 60c A 60 Bengal gram dal 7C, 16C J,M 12 Butter beans 39C, 52C J,A 46

DIABETES CARE, VOL. 11, NO. 2, FEBRUARY 1988 151 GLYCEMIC INDEX

TABLE 1 (Continued)

Food Gl Values* Subjectst Mean G

Legumes (Continued) Chick peas 46% 52C B,A 49 Green peas Dried 32", 68C C,A 50 Frozen 55", 74a C,A 65 Haricot (white) beans 44% 45% 56% 84 O,A,P,Q 57 ± 10 Kidney beans 27% 42% 65b M,A,B 45 ± 11 Red 25% 42% 43C S,A,B 37 ± 6b 10% 19C J,A 15 Soy beans Dried 20c A 20 Canned 22C A 22 Fruit Apple 48", 57C C,A 53 66", 81, 90, 99 C,B,A,J 84 ± 7 Orange 46% 58% 73" J,A,C 59 ± 8a Orange juice 67a A 67 Sugars 29% 30", 35b A,C,K 31 ± 2b Glucose 122% 131, 132", 137% E,K,R,C,D,J,A,M 138 ± 4C 137", 141% 145% 158a 126 A 126 b 152 A 152 85", 86, 91, 92 C,A,K,J 89 ± 2b Snack foods Corn chips 99" L 99 Potato chips 74% 79" A,L 77 Dairy products Ice cream 52b A 52 Skim milk 46C A 46 Whole milk 49c A 49 Yogurt 52C A 52 individual mean values from different groups of subjects. For significant difference from white bread (Gl = 100): aP < .05; bP < .01; CP < .001; "significance not given and unable to be calculated. tSubjects (refs.): A, normal English (11); B, predominantly NIDDM Canadian (13); C, diabetic German (4); D, normal American (6); E, NIDDM American (8); F, NIDDM Canadian (14); G, NIDDM and IDDM Canadian (20); H, NIDDM and IDDM Canadian (15); I, NIDDM and IDDM Canadian (16); J, normal rural African (18); K, NIDDM Canadian (19); L, normal Australian (20); M, normal Indian (21); N, normal American (22); O, normal Canadian (23); P, NIDDM Canadian (24); Q, IDDM Canadian (24); R, impaired GTT (7); S, normal Canadian (25). ^Significant difference from 100 given where >3 mean values are available. these foods was consistent for each individual (i.e., for those foods, we believe the Gl concept can be ap- whether for each patient the mean value for the low-GI plied to individual diets composed of many foods, foods was significantly below that of bread, a higher Gl Lack of agreement between different centers. Dis- food). For each subject, the mean Gl of the foods was similarities have been observed between the glycemic significantly below that of bread despite the wide vari- responses to certain foods tested in different centers, ation in individual responses to a given food. It cannot notably potato and rice (5-8,11,13). However, closer be inferred from these data whether certain individuals examination of the foods reveals that the center with consistently show lesser changes in Gl than others. It is the consistently higher glycemic response to potato also not apparent to what extent the differences from fed a 317-g baked russet potato (6,8), whereas the cen- the expected values are due to intraindividual variabil- ter with the consistently lower response fed a 273-g boiled ity; i.e., had each individual repeated each test on sev- new potato (11,13). The difference in weight fed, due eral occasions, the mean would probably more closely to the use of different food tables, accounts for part of approximate the expected Gl value (23; unpublished the difference in glycemic response. There may also be observations). Nevertheless, because each diabetic vol- true but unidentified differences between the more pow- unteer demonstrated a mean Gl value for the foods tested dery russet potato and the glutinous new potato. Ex- that was similar to or below the predicted mean Gl value amination of the types of rice fed indicate that the center

152 DIABETES CARE, VOL. 11, NO. 2, FEBRUARY 1988 D.J.A. JENKINS, T.M.S. WOLEVER, AND A.L. JENKINS

with the lower response (5-8) fed parboiled rice, whereas 300 the higher result was obtained in a center that fed regular rice (11,13). Subsequent testing has demonstrated that 250 =3 parboiled cereals, whether rice (16) or wheat (15), are 200 notable in resulting in relatively flat blood glucose mg/100 ml 150 profiles. Such differences are not simply due to lack of reproducibility but represent true differences in 100 physiologic effects between foods that previously were 50 considered the same. Other differences have also been 0 reported with respect to rice. Varieties of long-grain rice 60 120 180 may be higher in starch and consequently give Time (min) a flatter blood glucose response than the more amylo- pectin-rich short-grain varieties (44). FIG. 2. Mean plasma glucose levels of 8 NIDDM subjects In addition, the ripeness of fruits will determine their fed standard test meals containing baked potato (•—•), sugar content, a factor shown to be especially important rice (•—•), spaghetti (O—O), or (O—O) as major in the case of (45). Cooking will enhance the source of carbohydrate (34). degree of gelatinization of starch (46) and hence the degree to which it raises the blood glucose (47). These meal, differences in glycemic responses between the and many other food-related factors determining post- foods are abolished. Several studies fail to show any prandial glycemia and insulin response are emerging difference in glycemic response to mixed meals (48). (33,34,54,55). With the first study to apply Gl in this Substances such as phytates (49,50) and lectins (51) situation (Fig. 2), a major problem in interpretation was and indeed the content (52) are all known the use of total rather than incremental areas for com- to influence glycemic response and are altered by dif- parison of postprandial responses (56). ferent growing conditions (53). Finally, the absolute The suggestion that clinically there may be no great amounts fed by different investigators may depend on advantage from using the Gl to achieve a modest re- the food tables used or whether a direct analysis was duction in postprandial glycemia when the fasting blood performed. If so, the method used to determine dietary glucose value is grossly elevated is uncontested. The fiber content will influence the available carbohydrate primary concern must be the reduction of the fasting content. To some, the field might appear to be too vari- blood glucose level. However, if Gl is to be used to able to allow meaningful interpretation. An alternative rank the postprandial glucose responses to different view would be that much knowledge is being acquired meals, then the method of assessment would seem im- that will change our perception of food systems but will portant. The fasting blood glucose is not influenced by allow predictions to be made based on knowledge of the subsequent meal. However, if the total blood glu- physiologic responses to foods. A surprising fact is that, cose area is chosen, a large variation in starting value despite all these unknowns, there is a broad measure of could obscure differences between meal responses when agreement on the relative glycemic effect of many car- expressed as absolute postprandial levels. The Gl clas- bohydrate foods tested in different centers (Table 1). sification has therefore been based on incremental Lack of difference between mixed meals. The most responses. Similar treatment should be given if Gl is to topical criticism of the Gl concept is that, when indi- be used to predict the mixed-meal response. vidual carbohydrate foods are taken as part of a mixed In addition, if absolute values are used to calculate

TABLE 2 Variability of individual glycemic indexes (Gl) for 6 low-GI cereal foods compared with white bread tested in 8 diabetic patients

NIDDM patients IDDM patients

Food Gl 1 2 3 4 5 6 7 8 Mean ± SE

Bread 100 100 100 100 100 100 100 100 100 100.0 i 0 Bulgur 65 49 55 38 70 82 63 31 62 56.0 i 5.9 Pumpernickel 78 69 70 79 98 55 70 76 90 75.9 i 4.7 Parboiled rice 67 52 77 57 73 57 59 64 49 61.0 i 3.5 Barley 31 28 31 29 16 48 30 7 31 27.5 i 4.2 Wheat kernels 63 36 48 47 57 93 57 46 81 58.2 i 6.8 Rye kernels 47 31 41 27 43 51 59 33 76 45.1 i 5.8 Mean ± SE 58.5 ± 6.8 44.2 ± 6.3 53.7 ± 7.1 46.2 ± 8.0 59.5 ±11.4 64.3 ± 7.6 56.3 ± 5.6 42.8 ± 10.1 64.8 ± 9.0 53.9 i 5.8 Significance vs. bread (P) <.01 <.001 <.OO5 <.005 <.02 <.01 <.001 <.005 <.01 <.001

DIABETES CARE, VOL. 11, NO. 2, FEBRUARY 1988 153 GLYCEMIC INDEX

Gl, then the higher the fasting blood glucose value, the the validity of applying the Gl in this situation. Where smaller the contribution of the postprandial response to this has been done, a degree of predictability has been total glycemia. This can be demonstrated by analyzing found (Fig. 4; 58). data from a different study. Figure 3 illustrates both the There are at least five other published studies that absolute (top) and incremental {bottom) glycemic re- have examined the effects of mixed meals (32,33, sponses of 15 NIDDM subjects who ate meals of bread, 35,59,60). Only two of these are generally quoted (39). rice, spaghetti, and barley to which the same amount In one, it was concluded that the glycemic responses to of fat and protein as cheddar cheese had been added the meals "were similar except for one meal" (33). The (57). The incremental areas for rice, spaghetti, and bar- different meal (meal B) had a significantly greater gly- ley were 23, 44, and 59% less, respectively, than that cemic response than two of the other meals (meals A for bread, whereas the total areas for these meals were and C), as predicted by the Gl of meal B, and was 21 only 2, 19, and 25% less than that of bread. These and 27 Gl units greater than meals A and C (23). In reduced figures further diminished the chance of a re- addition, although Bantleetal. (32) concluded that fruc- lationship between the Gl for single foods and mixed tose-containing meals were not always lower than those meals (39). Furthermore, we consider it important that containing other , they were able to dem- the foods tested in a mixed meal should also have been onstrate a significant difference in NIDDM subjects (who tested singly before conclusions are drawn relating to are less variable than IDDM subjects) (23). In a later study by this group, small but significant differences be- tween meals of differing predicted glycemic effect were found in normal subjects, but they were not seen in 300 NIDDM subjects (54). Three less-known studies show good predictive ability with Gl. Parillo et al. (59) found the expected difference between bread and spaghetti when incorporated into a mixed meal (Fig. 5). Slama et al. (60,61) also found that blood glucose and insulin responses for different foods 200 in a mixed meal ranked as expected. Finally, Collier et al. (35) fed five different mixed meals to NIDDM sub- jects (Fig. 6). These resulted in a range of differences of almost 100 mg/dl in postprandial blood glucose levels. The close correlation between the expected Gl of the 100 meals and the observed glycemic responses was prob- ably a reflection of the fact that the foods fed had been tested previously and were known to have glycemic re- 0) sponses equivalent to their published Gl values (Fig. 7). o Therefore, before concluding that differences in gly- cemic response to individual foods are lost when they 0 are combined in a mixed meal, it appears important to pretest the individual carbohydrate components of the 200 mixed meal. O o EFFECTS OF DIETARY CHANGE 100 Only two studies have been published of the effect of incorporating carbohydrate foods that cause relatively low rises in blood glucose into the diet. In one, diabetic children reduced the glycemic impact of their diets for 6 wk by eating carbohydrate foods known to raise blood glucose minimally instead of more conventional car- bohydrate foods (67). This change resulted in improved 0 1 2 3 glucose tolerance and a fall in serum cholesterol after a Time (hours) standard meal. Although no significant fall was seen in HbA levels, there was a significant fall in glycosylated FIG. 3. Mean blood glucose concentrations {top) and blood 1c glucose increments {bottom) of 15 NIDDM subjects fed albumin, probably due to the much shorter half-life of test meals containing 50 g carbohydrate from white bread this protein, which makes it a more suitable marker for (A), polished rice (O), spaghetti (•), or pearled barley (O). relatively short dietary studies (63). In the second study To each food, 32 g cheddar cheese and 100 g cooked to- the same dietary maneuver was undertaken by a group mato were added (57). of hypertriglyceridemic, predominately glucose-intol-

154 DIABETES CARE, VOL. 11, NO. 2, FEBRUARY 1988 D.J.A. JENKINS, T.M.S. WOLEVER, AND A.L. JENKINS

350

300 I 250 I 200

200

0

FIG. 4. Mean plasma glucose and insu- 60 PLASMA INSULIN lin levels of NIDDM subjects fed test meals 50 containing white beans processed in 2 40 different ways: O, A, damaged cell walls; 30 •, undamaged cell walls. Left panel, re- sponses to bean products fed alone. Right 20 panel, responses when bean products were consumed as part of mixed meal 30 60 120 180 (58). *P < .05, **P < .01. TIME Imin) erant individuals over 1-mo periods (14). This dietary The conclusion is supported by some of the most suc- exchange was accompanied by falls in serum triglyc- cessful dietary studies to show improvement in glucose eride and total and low-density lipoprotein cholesterol control in diabetic patients (64-69) or reductions in blood levels. In these studies, there were small to modest in- lipids in hyperlipidemic individuals (66,70). Although creases in dietary fiber, which seemed unlikely to pro- the thrust of such studies was to increase fiber intake, it vide the whole explanation. was achieved with foods with a lesser impact on blood glucose than many of the foods they displaced. Dietary- fiber studies where this has not been the case have re- A Blood glucose (mmol/l)

Q 300 CD

GO CD 200

(JD

CD CD 100 —J PQ

0 12 3

• ' L. TIME (HOURS) 0 30 60 90 120 150 180 210 240 270 300 Time (min) FIG. 6. Mean blood glucose responses of 6 NIDDM sub- jects fed mixed meals containing instant potato (•), white FIG. 5. Mean ± SE blood glucose increments of 7 diabetic bread (A), polished rice (Q), white spaghetti (D), or mix- subjects fed mixed meals containing white bread (O), new ture of red lentils and barley (O) as major carbohydrate potato (•), or spaghetti (•) as major carbohydrate source source. Values that differ by >37 mg/dl are significantly (59). *P < .05, **P < .025, ***P < .01 vs. spaghetti. different. P < .05 (35).

DIABETES CARE, VOL. 11, NO. 2, FEBRUARY 1988 155 GLYCEMIC INDEX

The Gl approach to classification has also highlighted the urgent need for uniform food tables giving true avail- able or absorbable carbohydrate (starch and sugars) and dietary fiber separately so that known amounts of car- bohydrate are fed. Different portion sizes are used by different investigators, and differences in results may simply be due to the amounts of food fed. 1OOO APPLICATION OF GLYCEMIC INDEX

Strict application of Gl exchange principles is only pos- UJ sible in a research setting, where the Gl of diets may be calculated (23,56). After standard dietary advice, the control or "normal" diet for most diabetic and hyper- UJ 500 lipidemic patients has a Gl of 85-90 (14,63,76) (white CO bread = 100). This may be reduced by a mean 11-13 © (14,62). These changes are not large numerically but require a considerable change in the nature of the car- bohydrate foods eaten (Fig. 8). Thus, lower-GI foods such as pumpernickel bread will be increased from vir- C5 tually nonexistent levels, and regular wheat breads will be substantially reduced (Fig. 8). Nevertheless, for those not in a research setting, it may be possible to make use 20 40 60 80 100 MEAL G.I. FIG. 7. Correlation between mean incremental blood glu- RYE cose response areas for meals illustrated in Fig. 6 with BREAD expected meal glycemic index. P < .01, r = .9875. 20 OAT BRAN BULGUR BEANS BARLEY SPAGHETTI suited in much lesser benefits (71-75), and the benefits seen in lipid and carbohydrate metabolism may have been largely attributable to the accompanying reduction 0 in the proportion of fat in the diet. 40

UNANSWERED QUESTIONS

The major question that still remains is: what clinical 20 BREAKFAST gains can be expected through tighter control of post- POTATO RICE CEREALS prandial glycemic excursions? This can be resolved by dietary trials where low-GI foods are fed. However, if these trials are to be undertaken with the necessary de- gree of compliance, then an expanded list of classified foods is required. FRUIT Gl studies have drawn attention to the agreement and 20 BAKED disagreement between investigators in terms of the blood GOODS SUCROSE glucose responses observed after both single foods and mixed meals. The similarities provide hope that a system of classification, if comprehensive, may be feasible, useful as an adjunct to food tables in designing therapeutic diets, and a stimulus to studies of mechanism. The dis- FIG. 8. Intakes of different carbohydrate foods expressed agreements indicate there is much more to learn in terms as proportion of total dietary carbohydrate. Open bars, of processing and the effect of food components and control; solid bars, periods of low glycemic index (Gl). food form on physiological processes (e.g., digestion, Upper panel, foods that increased during low Gl; middle transit time, endocrine responses) before the most effec- panel, foods that decreased during low Gl; lower panel, tive use can be made of the knowledge. foods that remained unchanged during low Gl.

156 DIABETES CARE, VOL. 11, NO. 2, FEBRUARY 1988 D.J.A. JENKINS, T.M.S. WOLEVER, AND A.L. JENKINS

of Gl data by selecting foods to incorporate into patients' 4. Otto H, Niklas L: Differences d'action sur la glycemie diets that have the desired nutritional profile, complying d'aliments contenant des hydrates de carbone: conse- with current guidelines, and yet have a lower glycemic quences pour le traitment dietetique du diabete sucre. impact. Many of these are traditional or ethnic foods, Med Hyg 38:3424-29, 1980 e.g., pasta, lentils, beans, parboiled rice, barley, bulgur, 5. Crapo PA, Reaven G, Olefsky J: Plasma glucose and in- sulin responses to orally administered simple and com- and pumpernickel bread, which, rather than constrict- plex carbohydrates. Diabetes 25:741-47, 1976 ing the patients' eating habits, may in effect introduce 6. Crapo PA, Reaven G, Olefsky J: Postprandial plasma-glu- the patients to new foods. cose and -insulin responses to different complex carbo- Current recommendations by several agencies con- hydrates. Diabetes 26:1178-83, 1977 cerned with health (including heart foundations and 7. Crapo PA, Kolterman OG, Waldeck N, Reaven GM, Olef- cancer institutes in addition to diabetes associations) sky JM: Postprandial hormonal responses to different types support the increased use of carbohydrate foods. Over of complex carbohydrate in individuals with impaired the last decade, the overall aim has been to reduce con- glucose tolerance. Am ) Clin Nutr 33:1723-28, 1980 sumption of saturated fat, which is implicated in raising 8. Crapo PA, Insel J, Sperling M, Kolterman OG: Compari- serum cholesterol levels, and total dietary fat, which is son of serum glucose, insulin, and glucagon responses to different types of complex carbohydrate in non-insulin- associated epidemiologically with colon and breast tu- dependent diabetic patients. Am J Clin Nutr 34:184-90, mors. It has not been suggested that the increased car- 1981 bohydrate that replaces fat should come from sugars. 9. Jenkins DJA, Ghafari H, WoleverTMS, Taylor RH, Barker Nevertheless sugar may not raise the blood glucose HM, Fielden H, Jenkins AL, Bowling AC: Relationship more than many starchy foods (32), and fructose sub- between the rate of digestion of foods and post-prandial stitution actually results in significantly flatter postpran- glycemia. Diabetologia 22:450-55, 1982 dial glucose responses (77). Therefore, modest amounts 10. Jenkins DJA, Wolever TMS, Thome MJ, Jenkins AL, Wong of sugars may be used as sweeteners. However, their GS, Josse RG, Csima A: The relationship between gly- use as a source of calories is still a matter of debate. cemic response, digestibility, and factors influencing the There is concern that in susceptible individuals, fructose dietary habits of diabetics. Am I Clin Nutr 40:1175-91, may raise serum triglyceride levels (78,79). Further- 1984 11. Jenkins DJA, Wolever TMS, Taylor RH, Barker HM, Fiel- more, when sucrose replaces starch in diets higher in den H, Baldwin JM, Bowling AC, Newman HC, Jenkins saturated fats or very high in carbohydrate, it may in- AL, Goff DV: Glycemic index of foods: a physiological crease the levels of both cholesterol and triglyceride basis for carbohydrate exchange. Am) Clin Nutr 34:362- (80,81). Thus, despite the fact that high fat or fructose- 66, 1981 containing foods may cause relatively flat blood glucose 12. Jenkins DJA, Wolever TMS, Jenkins AL, Josse RG, Wong responses, they cannot be recommended solely on the GS: The glycemic response to carbohydrate foods. Lancet basis of their lower acute glycemic response. The Gl 2:388-91, 1984 classification may therefore be most appropriately used 13. Jenkins DJA, Wolever TMS, Jenkins AL, Thome MJ, Lee to rank starchy foods. These starchy foods would already R, Kalmusky J, Reichert R, Wong GS: The glycemic index have been chosen for possible inclusion in the diet on of foods tested in diabetic patients: a new basis for car- bohydrate exchange favouring the use of legumes. Dia- the basis of their nutritional attributes. In this setting the betologia 24:257-64, 1983 Gl would allow selection of foods that have the added 14. Jenkins DJA, WoleverTMS, Kalmusky J, Giudici S, Gior- advantage of producing lower postprandial glycemic ex- dano C, Wong GS, Bird JH, Patten R, Hall M, Buckley cursions. GC, Little JA: Low glycemic index foods in the man- agement of hyperlipidemia. Am J Clin Nutr 42:604-17, 1985 ACKNOWLEDGMENTS 15. Jenkins DJA, Wolever TMS, Jenkins AL, Giordano C, Giu- dici S, Thompson LU, Kalmusky J, Josse RG, Wong GS: These studies were supported by the National Sci- Low glycemic response to traditionally processed wheat and rye products: bulgur and pumpernickel bread. Am I ences and Engineering Research Council, Canada. Clin Nutr 43:516-20, 1986 16. Wolever TMS, Jenkins DJA, Kalmusky J, Jenkins AL, Gior- dano C, Giudici S, Josse RG, Wong GS: Comparison of REFERENCES regular and parboiled : explanation of discrepancies between reported glycemic responses to rice. Nutr Res 1. Wagner R, Warkany J: Untersuchungen liber den zuck- 6:349-57, 1986 erbildenden Wert der Gemuse in der Diabetikerkost (Ab- 17. Wolever TMS, Jenkins DJA, Kalmusky J, Giordano C, Giu- stract). Z Kinderheilkd 44:322, 1927 dici S, Jenkins AL, Thompson LU, Wong GS, Josse RG: 2. Conn JW, Newburgh LH: The glycemic response to iso- Glycemic response to pasta: effect of food form, cooking glucogenic quantities of protein and carbohydrate. / Clin and protein enrichment. Diabetes Care 9:401-404, 1986 Invest 15:665-71, 1936 18. Walker ARP, Walker BR: Glycemic index of South African 3. Otto H, Bleyer G, Pennartz M, Sabin G, Schauberger G, foods determined in rural blacks—a population at low Spaethe K: Kohlenhydrataustausch nach biologischen risk to diabetes. Hum Nutr Clin Nutr 36:215-22, 1984 aquivalenten. In Diatetik bei Diabetes Mellitus. Otto H, 19. Wolever TMS, Wong GS, Kenshole A, Josse RG, Thomp- Spaethe R, Eds. Bern, Huber, 1973, p. 41-50 son LU, Lam KY, Jenkins DJA: in the diabetic diet:

DIABETES CARE, VOL. 11, NO. 2, FEBRUARY 1988 157 GLYCEMIC INDEX

a comparison with other carbohydrates. Nutr Res 5:1335- 35. Collier GR, Wolever TMS, WongGS, Josse RG: Prediction 45, 1985 of glycemic response to mixed meals in non-insulin-de- 20. Brand JC, Nicholson PL, Thorburn AW, Truswell AL: Food pendent diabetic subjects. Am j Clin Nutr 44:349-52, processing and the glycemic index. Am J Clin Nutr 1986 42:1192-96, 1985 36. Jenkins DJA, Wolever TMS, Jenkins AL, Lee R, Wong GS, 21. Dilawari JB, Kamath PS, Batta RP, Mukewar S, Raghavan Josse R: Glycemic response to wheat products: reduced S: Reduction of postprandial plasma glucose by bengal response to pasta but no effect of fiber. Diabetes Care gram dal (Cicer arietnum) and rajmah (Phaseolus vul- 6:155-59, 1983 garis). Am j Clin Nutr 34:2450-53, 1981 37. Jenkins DJA, Wolever TMS, Taylor RH, Barker H, Fielden 22. Potter JG, Coffman KP, Reid RL, Drall JM, Albrink MJ: H: Exceptionally low blood glucose response to dried Effect of test meals of varying dietary fiber content on beans: comparison with other carbohydrate foods. Br Med plasma insulin and glucose response. Am j Clin Nutr I 2:578-80, 1980 34:328-34, 1981 38. Coulston AM, Hollenbeck CB, Reaven GM: Utility of 23. Wolever TMS, Nuttall FQ, Lee R, Wong GS, Josse RG, studies measuring glucose and insulin responses to var- Csima A, Jenkins DJA: Prediction of the relative blood ious carbohydrate-containing foods. Am } Clin Nutr 39: glucose response of mixed meals using the white bread 163-65, 1984 glycemic index. Diabetes Care 8:418-28, 1985 39. Hollenbeck CB, Coulston AM, Reaven GM: Glycemic 24. Jenkins DJA, Wolever TMS, Wong GS, Kenshole A, Josse effects of carbohydrates: a different perspective. Diabetes RG, Thompson LU, Lam KY: Glycemic responses to foods: Care 9:641-47, 1986 possible differences between insulin-dependent and non- 40. Kolata G: Diabetics should lose weight, avoid fad diets. insulin-dependent diabetics. Am ) Clin Nutr 40:971-81, Science 235:163-64, 1987 1984 41. Simpson RW, McDonald J, Wahlqvist ML, Atley L, Outch 25. Wolever TMS, Cohen Z, Thompson LU, Thorne MJ, Jen- K: Food physical factors have different effects in non- kins MJA, Prokipchuk EJ, Jenkins DJA: Heal loss of avail- diabetics and diabetics. Am j Clin Nutr 42:462-69, 1985 able carbohydrate in man: comparison of a breath hydro- 42. Simpson RW, McDonald J, Wahlqvist ML, Atley L, Outch gen method with direct measurement using a human K: Macronutrients have different metabolic effects in non- ileostomy model. Am / Gastroenterol 81:115-22, 1986 diabetics and diabetics. Am J Clin Nutr 42:449-53, 1985 26. Vaaler S, Hanssen KF, Aagenaes O: Plasma glucose and 44. Goddard MS, Young G, Marcus R: The effect of amylose insulin responses to orally administered carbohydrate-rich content on insulin and glucose responses to ingested rice. foodstuffs. Nutr Metab 24:168-75, 1980 Am / Clin Nutr 39:388-92, 1984 27. Vaaler S, Wiseth R, Aagenaes O: Increase in blood glu- 45. Englist HN, Cummings JH: Digestion of the polysaccha- cose in insulin-dependent diabetics after intake of various rides of banana in the human small intestine (Abstract). fruits. Ada Med Scand 212:281-83, 1982 Int Congr Nutr, 13th, Brighton, UK, 1985, p. 70 28. Tappy L, Wursch P, Randin JP, Felber JP, Jequier E: Met- 46. Booher CE, Behan I, McNeans E: Biologic utilization of abolic effect of pre-cooked instant preparations of bean unmodified and modified food (Abstract), j Nutr and potato in normal and in diabetic subjects. Am J Clin 45:75, 1951 Nutr 43:30-36, 1986 47. Collings P, Williams C, MacDonald I: Effect of cooking 29. Jenkins DJA, Wolever TMS, Taylor RH, Barker HM, Fiel- on serum glucose and insulin responses to starch. Br Med den H, Jenkins AL: Effect of guar crispbread with cereal 7 282:1032-33, 1981 products and leguminous on blood glucose con- 48. Thorne MJ, Thompson LU, Jenkins DJA: Factors affecting centrations of diabetics. Br Med I 281:1248-50, 1980 starch digestibility and the glycemic response with special 30. Coulston AM, Greenfield MS, Enger F, Tokey T, Reaven reference to legumes. Am J Clin Nutr 38:481-88, 1983 GM: Effect of source of dietary carbohydrate on plasma 49. Yoon JH, Thompson LU, Jenkins DJA: The effect of phytic glucose and insulin responses to test meals in normal sub- acid on in vitro rate of starch digestibility and blood glu- jects. Am j Clin Nutr 33:1279-82, 1980 cose response. Am ) Clin Nutr 38:835-42, 1983 31. Coulston AM, Greenfield MS, Kraemer FB, Tobey TA, 50. Thompson LU, Yoon JH, Jenkins DJA, Wolever TMS, Reaven GM: Effect of differences in source of dietary car- Jenkins AL: Relationship between polyphenol intake and bohydrate on plasma glucose and insulin responses to blood glucose response of normal and diabetic individ- meals in patients with impaired carbohydrate tolerance. uals. Am ) Clin Nutr 39:745-51, 1984 Am I Clin Nutr 34:2716-20, 1981 51. Rea RL, Thompson LU, Jenkins DJA: Lectins in foods and 32. Bantle JP, Laine DC, Castle GW, Thomas JW, Hoogwerf their relation to starch digestibility. Nutr Res 5:919-29, BJ, Goetz FC: Postprandial glucose and insulin responses 1985 to meals containing different carbohydrates in normal and 52. Jenkins DJA, Wolever TMS, Leeds AR, Gassull MA, Di- diabetic subjects. N Engl J Med 309:7-12, 1983 lawari JB, Goff DV, Metz GL, Alberti KGMM: Dietary 33. Nuttall FQ, Mooradian AD, DeMarais R, Parker S: The fibers, fiber analogues and glucose tolerance: importance glycemic effect of different meals approximately isoca- of viscosity. Br Med) 1:1392-94, 1978 loric and similar in protein, carbohydrate, and fat content 53. Al-Nouri FF, Siddiqi AM: Biochemical evaluation of twelve as calculated using the ADA exchange lists. Diabetes Care broad bean cultivars. Can Inst Food Sci Technol) 15:37— 6:432-35, 1983 40, 1982 34. Coulston AM, Hollenbeck CB, Liu GC, William RA, Star- 54. Laine DC, Thomas JW, Bantle JP: Comparison of the pre- ich GH, Mazzaferri EL, Reaven GM: Effect of source of dictive capabilities of the diabetic exchange lists and the dietary carbohydrate on plasma glucose, insulin, and gas- glycemic indices of foods (Abstract). Diabetes 35 (Suppl. tric inhibitory polypeptide responses to test meals in sub- 1):43A, 1986 jects with non-insulin-dependent diabetes mellitus. Am ) 55. Coulston A, Hollenbeck C: Comparison of plasma glu- Clin Nutr 40:965-70, 1984 cose and insulin responses to mixed meals of predicted

158 DIABETES CARE, VOL. 11, NO. 2, FEBRUARY 1988 D.J.A. JENKINS, T.M.S. WOLEVER, AND A.L. JENKINS

high, medium, and low glycemic response (Abstract). Di- 68. Simpson HRC, Simpson RW, Lousley S, Carter RD, Geekie abetes 35 (Suppl. 1):43A, 1986 M, Hockaday TDR, Mann Jl: A high carbohydrate le- 56. Wolever TMS, Jenkins DJA: The use of the glycemic index guminous fibre diet improves all aspects of diabetic con- in predicting the blood glucose response to mixed meals. trol. Lancet 1:1-5, 1981 Am ) Clin Nutr 43:167-72, 1986 69. Rivellese A, Riccardi G, Giacco A, Pancioni D, Genovese 57. Wolever TMS, Jenkins DJA, Josse RG, Wong GS, Lee R: S, Mattioli PL, Mancini M: Effect of dietary fiber on glu- The glycemic index: similarity of values derived in insu- cose control and serum lipoproteins in diabetic patients. lin-dependent and non-insulin-dependent diabetic pa- Lancet 2:447-50, 1980 tients. I Am Col Nutr 6:295-305, 1987 70. Anderson JW, Chen WJL, Sieling B: Hypolipidemic effects 58. Golay A, Coulston AM, Hollenbeck CB, Kaiser LL, Wursch of high carbohydrate, high fiber diets (Abstract). Metab- P, Reaven GM: Comparison of metabolic effects of white olism 29:551, 1980 beans processed into two different physical forms. Dia- 71. Simpson HCR, Carter RD, Lousley S, Mann Jl: Digestible betes Care 9:260-66, 1986 carbohydrate—an independent effect on diabetic control 59. Parillo M, Giacco R, Riccardi G, Pacioni C, Rivellese in type 2 (non-insulin-dependent) diabetic patients? Dia- A: Different glycaemic responses to pasta, bread, and betologia 23:235-39, 1982 potatoes in diabetic patients. Diabetes Med 2:374-77, 72. Manhire A, Henry CL, Hartog M, Heaton KW: Unrefined 1985 carbohydrate and dietary fiber in treatment of diabetes 60. Slama G, Bornet F, Blayo A, Costagliola D, Haardt MJ, mellitus. I Hum Nutr 35:99-101, 1981 Tchobroutsky G: Insulinogenic and glycaemic indexes of 73. Hollenbeck CB, Riddle MC, Connor WE, Leklem JE: The various starch-rich foods taken in a mixed meal or alone effects of subject-selected high carbohydrate, low fat diets by type 2 diabetics (Abstract). Diabetes 34 (Suppl. 1 ):48A, on glycemic control in insulin dependent diabetes mel- 1985 litus. Am I Clin Nutr 41:293-99, 1985 61. Bornet FRJ, Costagliola D, Blayo A, Fontvieille A, Haardt 74. Hollenbeck CB, Coulston AM, Reaven GM: To what ex- MJ, Letanoux M, Tchobroutsky G, Slama G: Insulinogenic tent does increased dietary fiber improve glucose and lipid and glycemic indices of six starch-rich foods taken alone metabolism in patients with non-insulin-dependent dia- and in a mixed meal by type 2 diabetics. Am j Clin Nutr betes mellitus (NIDDM)? Ami Clin Nutr 43:16-24, 1986 45:588-95, 1987 75. Lindsay AN, Hardy S, Jarrett L, Rallinson ML: High-car- 62. Collier GR, Kalmusky J, Giudici S, Helman G, Giordano bohydrate, high-fiber diet in children with type I diabetes C, Ehrlich RM: Effects of slowly digested carbohydrates mellitus. Diabetes Care 7:63-67, 1984 in type I diabetic children (Abstract). Diabetes 34 (Suppl. 76. Wolever TMS, Jenkins DJA: Application of the glycaemic 1):33A, 1985 index to mixed meals (Letter). Lancet 2:944, 1985 63. Jones IR, Owens DR, Williams S, Ryder REJ, Birtwell AJ, 77. Crapo PA, Scarlett JA, Kolterman OG: Comparison of the Jones MK, Gicheru K, Hayes TM: Glycosylated serum metabolic responses to fructose and sucrose in sweetened albumin: an intermediate index of diabetic control. Dia- foods. Am ) Clin Nutr 36:256-61, 1982 betes Care 6:501-503, 1983 78. Crapo PA, Kolterman OG, Henry RR: Metabolic conse- 64. Kiehm TG, Anderson JW, Ward K: Beneficial effects of a quences of two-week fructose feeding in diabetic sub- high carbohydrate high fiber diet in hyperglycemic men. jects. Diabetes Care 9:111-19, 1986 Am J Clin Nutr 29:895-99, 1976 79. Hallfrizch J, Reiser S, Prather ES: Blood lipid distribution 65. Anderson JW, Ward K: High carbohydrate, high fiber diets of hyperinsulinemic men consuming three levels of fruc- for insulin treated men with diabetes mellitus. Am j Clin tose. Am I Clin Nutr 37:740-48, 1983 Nutr 32:2312-21, 1979 80. Antar MA, Little JA, Lucas C, Buckley GC, Csima A: Inter- 66. Anderson JW, Chen WL: Plant fiber: carbohydrate and relationship between the kinds of dietary carbohydrate lipid metabolism. Am J Clin Nutr 32:346-63, 1979 and fat in hyperlipoproteinemic patients. 67. Kinmonth AL, Angus RM, Jenkins PA, Smith MA, Baum 11:191-201, 1970 JD: Whole foods and increased dietary fiber improve blood 81. Albrink MJ, Ulrich IH: Interaction of dietary sucrose and glucose control in diabetic children. Arch Dis Child fiber on serum lipids in healthy young men fed carbo- 57:187-94, 1982 hydrate diets. Am ) Clin Nutr 43:417-28, 1986

DIABETES CARE, VOL. 11, NO. 2, FEBRUARY 1988 159