# An insulin index of foods: the insulin demand generated by 1000-kJ portions of common foods

> An annotated edition of this paper on Fermat's Library, where scientists and readers explain its ideas, equations and history in the margins. Annotated edition: https://fermatslibrary.com/s/an-insulin-index-of-foods-the-insulin-demand-generated-by-1000-kj-portions-of-common-foods

- **Authors:** Susanne HA Holt, Janette C Brand Miller, and Peter Petocz
- **Published:** 1997
- **Topic:** Biology
- **Original:** https://s3-us-west-1.amazonaws.com/fermatslibrary/uploads%2F1551203931223-hltjpwf3i3e-0c4bf227fe1d7ef6b1f4d82c7e8c9de7%2Finsulin_paper.pdf
- **Pages:** 13
- **Annotations:** 23 comments in 20 threads
- **Featured:** Fermat's Library Journal Club
- **Annotated edition:** https://fermatslibrary.com/s/an-insulin-index-of-foods-the-insulin-demand-generated-by-1000-kj-portions-of-common-foods

## Annotations

### Annotation 1 (page 1)

Beside the passage: “Susanne HA Holt, Janette C Brand Miller, and Peter Petocz Downloaded from https://academic.oup.com/ajcn/article-abstract/66/5/1264/4655967 by guest on 26 February 2019”

**Annotation** by Yaron Hadad (2019-03-01):

> Why should you read this paper? Glucose and insulin are two of the most important elements of metabolism. In an oversimplified nutshell - when we consume food our blood glucose usually increases. Consequently, the pancreas secretes insulin which lowers the glucose levels back to normal. Insulin is a hormone and is often considered to be the principal regulator of fat storage in the body (this is a bit controversial, but is mostly considered correct. Read the book "Why we get fat" by Gary Taubes for more info). Therefore knowing how the foods we eat spike our insulin levels can be helpful in determining what should be eaten for weight loss.

**Reply** by EVARISTO ZAMBRANO (2020-08-05):

> I AM DIABETIC

### Annotation 2 (page 1)

Beside the passage: “ABSTRACT The aim of this study was to systematically compare postprandial insulin responses to isoenergetic 1000-U (240-kcal) portions of several common foods. Correlations with nutrient content were determined. Thirty-eight foods separated into six food categories (fruit, bakery products, snacks, carbohydrate rich …”

**Annotation** by Yaron Hadad (2019-03-01):

> Postprandial = after the meal. This is usually measured within two hours after the meal.
>
> A word of caution: this paper averages postprandial insulin response over multiple subjects. However, in reality, different people react differently to different foods, so while the average values provided here can give you a general idea of what they usually do to people, it doesn't necessarily mean you will respond the same to those foods.

### Annotation 3 (page 1)

Beside the passage: “ABSTRACT The aim of this study was to systematically compare postprandial insulin responses to isoenergetic 1000-U (240-kcal) portions of several common foods. Correlations with nutrient content were determined. Thirty-eight foods separated into six food categories (fruit, bakery products, snacks, carbohydrate rich …”

**Annotation** by Yaron Hadad (2019-03-01):

> Glucose can be measured today using a CGM (continuous glucose monitor) that measures the glucose levels every few minutes. However, insulin is trickier and we must draw blood in order to measure it. So the people who participated in this study drew blood 8 times after each meal they ate!

### Annotation 4 (page 1)

Beside the passage: “ABSTRACT The aim of this study was to systematically compare postprandial insulin responses to isoenergetic 1000-U (240-kcal) portions of several common foods. Correlations with nutrient content were determined. Thirty-eight foods separated into six food categories (fruit, bakery products, snacks, carbohydrate rich …”

**Annotation** by Yaron Hadad (2019-03-01):

> Since an increase in glucose levels usually results in insulin secretion, there is a strong correlation between foods that are high in carbs (carbs can be easily converted to glucose and thus often cause a glucose spike) and foods that had high insulin spikes.

### Annotation 5 (page 1)

Beside the passage: “ABSTRACT The aim of this study was to systematically compare postprandial insulin responses to isoenergetic 1000-U (240-kcal) portions of several common foods. Correlations with nutrient content were determined. Thirty-eight foods separated into six food categories (fruit, bakery products, snacks, carbohydrate rich …”

**Annotation** by Yaron Hadad (2019-03-01):

> This was the most interesting insight of this paper in my opinion. There are foods that are relatively low in carbs (meats) and foods that are high in carbs (donuts and the like) that still cause an insulin spike despite not causing a significant glucose spike.
>
> The reason this is so interesting is that it shows the fundamental limitation glucose monitoring has. On one hand, glucose is the most interesting biomarker we can probably measure continuously today as it is directly related to our metabolism. It is especially interesting because it is the best proxy we have for insulin in the body. On the other hand, this result shows that this proxy is not perfect and has limitations.

### Annotation 6 (page 3)

Beside the passage: “Thus, postprandial insulin responses are not always propor tional to blood glucose concentrations or to a meal's total carbohydrate content. Several insulinotropic factors are known to potentiate the stimulatory effect of glucose and mediate postprandial insulin secretion. These include fructose, certain amino acids …”

**Annotation** by Yaron Hadad (2019-03-01):

> Protein is made of amino acids. This implies protein (or some forms of protein) can also cause an insulin spike.

### Annotation 7 (page 4)

Beside the passage: “completed two or more food categories. All of the subjects approved by the Medical Ethical Review Committee of the were university students; relevant characteristics ofthe subjects University of Sydney. are listed in Table 3. The mean body mass index (BMI, in kg/m2) of the 41 subjects was 22.7 Â±0.4 (range: …”

**Annotation** by Yaron Hadad (2019-03-01):

> Most of the subjects were healthy lean people without diabetes

### Annotation 8 (page 4)

Beside the passage: “completed two or more food categories. All of the subjects approved by the Medical Ethical Review Committee of the were university students; relevant characteristics ofthe subjects University of Sydney. are listed in Table 3. The mean body mass index (BMI, in kg/m2) of the 41 subjects was 22.7 Â±0.4 (range: …”

**Annotation** by Yaron Hadad (2019-03-01):

> The convention is to use white bread or sugar when computing glycemic index (this means that a food that has GI of 100% has the same spike as bread. Saying it has the same spike doesn't mean it has the same curve, the technical way to do it is to calculate the area under the glucose curve).
>
> So in this case they do the same with the insulin curve.

### Annotation 9 (page 5)

Beside the passage: “Cumulative changes in postprandial plasma glucose and insulin responses for each food were quantified as the incre mental area under the 120-mn response curve (AUC), which was calculated by using the trapezoidal rule with fasting con centrations as the baseline and truncated at zero. Any negative areas tended to be …”

**Annotation** by Yaron Hadad (2019-03-01):

> This is the best way to calculate an integral =)
>
> https://www.wikiwand.com/en/Trapezoidal_rule

### Annotation 10 (page 5)

Beside the passage: “Subjects were fed 1000-U portions of the test foods in a random order on separate mornings after a 10-h overnight fast. Within each food group, each subject acted as his or her own control, being tested at the same time of day and under as similar conditions as possible. Subjects were asked to refrain from unusual …”

**Annotation** by Yaron Hadad (2019-03-01):

> Glucose is very much affected by the food you ate, but also from the quality of sleep, exercise and stress levels. So it's common to try and clear potential noises by measuring the response to a meal after they fast enough (10 hours) at night and eat the meals immediately in the morning.

### Annotation 11 (page 5)

Beside the passage: “Finger-prick blood samples (1.5â€”2.5mL) were collected from warmed hands immediately before the meal (0 mm) and 15, 30, 45, 60, 75, 90, 105, and 120 mm after the start of the meal (into plastic tubes that had been kept on ice) with use of an automatic lancet device (Autoclix; Boehringer Mannheim Australia, Castle …”

**Annotation** by Yaron Hadad (2019-03-01):

> Ouch

### Annotation 12 (page 6)

Beside the passage: “fries70 Â±117643 Â±713146 Â±29209 Â±1982 Â±871 Â±1674 Â±12Whiterice129Â±168143Â±68369Â±5145Â±1240Â±3110Â±1579Â±12Whole-meal bread106 Â±141 1 203 Â±1420122 Â±20247 Â±311 11 Â±1497 Â±1796 Â±12Potatoes148Â±2413930Â±1467120Â±19284Â±3038Â±4141Â±35121Â±11Groupmeanâ€”8410Â±461106Â±8182Â±1062Â±588Â±674Â±8Protein-rich”

**Annotation** by Shital Shah (2019-03-12):

> There is not much difference between white or brown rice. White bread is practically evil, even worse than doughnuts. If you must eat cereals, make it All-Bran. Cheese and peanuts are great.

**Reply** by Yaron Hadad (2019-03-12):

> In glucose levels, but there is a big difference in insulin levels...

### Annotation 13 (page 7)

Beside the passage: “There were large differences in mean glycemic and insulin responses to the foods, both within and between food groups. Mean glucose and insulin AUC results, mean GSs and ISs, and the mean ratios of insulin to glucose AUCs (the amount of insulin secretion in relation to the blood glucose response) are listed in Table …”

**Annotation** by Yaron Hadad (2019-03-01):

> AUC = Area Under the Curve

### Annotation 14 (page 7)

Beside the passage: “Within each food group, the subjects' average fasting plasma glucose and insulin concentrations were not signif icantly different among the foods. Mean fasting plasma glucose concentrations did not vary significantly among the six food groups, whereas mean fasting insulin concentra tions were more variable, ranging …”

**Annotation** by Yaron Hadad (2019-03-01):

> They add it as a meaningless comment but there may be something much deeper here.

**Reply** by Brit (2019-03-19):

> There must be some published work on the relationship between insulin sensitivity (which is essentially mean fasting insulin) and BMI.

### Annotation 15 (page 7)

Beside the passage: “Overall, among the 38 test foods, jellybeans produced the highest mean IS (160 Â±16%), eightfold higher than the lowest IS (for peanuts: 20 Â±5%) (Figure 2). White bread, the stan dard food, consistently produced one of the highest glucose and insulin responses (peak and AUC) and had a higher IS than most of the other …”

**Annotation** by Yaron Hadad (2019-03-01):

> Don't eat jollybeans LOL!

### Annotation 16 (page 9)

Beside the passage: “+ 1.29 sugar + 0.377 starch (3)”

**Annotation** by Yaron Hadad (2019-03-01):

> These equations give the fit between glucose and insulin scores. Don't forget that there are real exceptions there (see the graph on the left).

### Annotation 17 (page 9)

Beside the passage: “Glucose score (%)”

**Annotation** by Yaron Hadad (2019-03-01):

> This graph is great and while it shows that there is a great correlation between glucose and insulin scores (thus - glucose is an excellent proxy to insulin) we see that it is not a perfect fit and there are several exceptions. The exceptions are fat & carb reach foods (e.g. donuts and croissants) and various meats.

### Annotation 18 (page 11)

Beside the passage: “FIGURE 5. Relations between the nutrient contents of the test foods and the mean insulin scores. Fiber: r = â€”¿0.10,NS, n = 32; protein: r = â€”¿0.24, NS,n = 38; totalcarbohydrate:r = 0.39,P < 0.05,n = 36; sugar:r = 0.36,P < 0.05,n 36; starch:r = â€”¿0.09,NS,n = 30; fat: r = â€”¿0.27,P < 0.05, n = 36.”

**Annotation** by Yaron Hadad (2019-03-01):

> This is a great figure. Together with the previous learnings we can conclude:
> - Glucose is an excellent proxy to insulin (but it is not a perfect proxy as we saw there are exceptions)
> - Carbs are the best proxy for glucose (again, far from being perfect). Another good proxy is sugars.
>
> However, concluding that carbs => insulin spikes is a huge leap as we accumulate the errors there. Moreover, another thing that was not discussed here is the individual responses people seem to have for meals.
>
> So this paper can just give us a rule of thumbs to work with, but without measuring your own body you can never know what foods would be best for you (at least in respect to glucose or insulin management).

### Annotation 19 (page 12)

Beside the passage: “Our study was undertaken to test the hypothesis that the postprandial insulin response was not necessarily proportional to the blood glucose response and that nutrients other than carbohydrate influence the overall level of insulinemia. Multi pIe-regression analysis of the individual results showed that the glycemic …”

**Annotation** by Yaron Hadad (2019-03-01):

> Glucose is great but is still only a limited proxy. That being said, if one is living with diabetes managing glucose can have very meaningful health implications (so the conclusion here should be taken with a grain of salt).

### Annotation 20 (page 12)

Beside the passage: “Our study was undertaken to test the hypothesis that the postprandial insulin response was not necessarily proportional to the blood glucose response and that nutrients other than carbohydrate influence the overall level of insulinemia. Multi pIe-regression analysis of the individual results showed that the glycemic …”

**Annotation** by EVARISTO ZAMBRANO (2020-08-05):

> NOT QUESTION

## Full text

Extracted from the PDF. [Annotation N] marks where each annotation above sits.

#### Page 1

ABSTRACT The aim of this study was to systematically [Annotation 2] compare postprandial insulin responses to isoenergetic 1000-U (240-kcal) portions of several common foods. Correlations with nutrient content were determined. Thirty-eight foods separated into six food categories (fruit, bakery products, snacks, carbohydrate rich foods, protein-rich foods, and breakfast cereals) were fed to groups of 11â€”13healthy subjects. Finger-prick blood samples were [Annotation 3] obtained every 15 mm over 120 mm. An insulin score was calcu lated from the area under the insulin response curve for each food with use of white bread as the reference food (score = 100%). Significant differences in insulin score were found both within and among the food categories and also among foods containing a similar amount of carbohydrate. Overall, glucose and insulin scores were highly correlated (r = 0.70, P < 0.001, n = 38). However, protein-rich foods and bakery products (rich in fat and [Annotation 4] refined carbohydrate) elicited insulin responses that were dispro portionately higher than their glycemic responses. Total carbohy [Annotation 5] drate (r = 0.39, P < 0.05, n = 36) and sugar (r = 0.36, P < 0.05, n = 36) contents were positively related to the mean insulin scores, whereas fat (r â€”¿0.27,NS, n 36) and protein (r â€”¿0.24,NS, n = 38) contents were negatively related. Consideration of insulin scores may be relevant to the dietary management and pathogen esis of non-insulin-dependent diabetes mellitus and hyperlipidemia and may help increase the accuracy of estimating preprandial insulin requirements. Am J Clin Nutr l997;66:l264â€”76.

KEY WORDS Insulin, glycemic index, NIDDM, non insulin-dependent diabetes meffitus, diabetic diet, hyperlipid emia, carbohydrate, insulin score, glucose score, area under the curve, humans

INTRODUCTION

The insulinemic effects of foods may be relevant to the treatment and prevention of weight gain, non-insulin-depen dent diabetes mellitus (NIDDM), and associated complications. Recent studies have shown that carbohydrate-rich diets, which result in high postprandial glucose and insulin responses, are associated with undesirable lipid profiles (1, 2), greater body fat (3â€”5),and the development of insulin resistance in rats (6) and humans (7, 8). Both obesity and NJDDM are associated with varying degrees of insulin resistance and fasting hyperin sulinemia. Prolonged or high degrees of postprandial insuline mia are thought to contribute to the development of insulin resistance and associated diseases (9â€”17).Therefore, the clas

sification of the relative insulinemic effects of different foods is of both theoretical and practical significance.

Postprandial blood glucose responses have been the focus of much research because of their importance for glycemic con trol in patients with diabetes. It is now well accepted that different foods containing equal amounts of carbohydrate can produce a wide range of blood glucose responses. The glyce mic index (GI) method was developed to rank foods according to the extent to which they increase blood glucose concentra tions (18). Tables of GI values of common carbohydrate containing foods are a useful guide to help people with diabetes choose foods that produce smaller glycemic responses. How ever, the GI concept does not consider concurrent insulin responses and few studies have reported GI values and their accompanying insulin responses.

The extent to which different dietary factors affect post prandial insulinemia has not been well researched because insulin secretion is largely assumed to be proportional to postprandial glycemia. Furthermore, hyperglycemia is thought to be more relevant to the secondary complications of NIDDM because the abnormal insulin secretion or action in people with diabetes is controlled with exogenous insulin or medications that counteract insulin resistance. However, knowledge of factors that influence both postprandial gly cemia and insulin secretion in nondiabetic persons is re quired to devise treatment strategies that will completely normalize meal-related glycemia (19).

To explore the importance of dietary habits and postprandial insulinemia in the etiology and treatment of NIDDM, we need to be able to systematically rate insulin responses to common foods. If we are to compare insulin responses to foods, what is the best basis of comparison? Should we compare insulin responses to portions of food representing a normal serving size, portions containing an equal amount of carbohydrate, or portions containing an equal amount of energy? 01 tables represent the glycemic effects of equal-carbohydrate portions

I From the Human Nutrition Unit, Department of Biochemistry, The University of Sydney; and the School of Mathematical Sciences, The University of Technology, Sydney, Australia.

2 Supported by research grants from The University of Sydney and Kellogg's Australia Pty Ltd.

3 Address reprint requests to JC Brand Miller, Human Nutrition Unit, Department of Biochemistry 008, The University of Sydney, NSW 2006, Australia.

Received November 21, 1996. Accepted for publication May 22, 1997.

1264 Am J Clin Nutr 1997;66:1264â€”76.Printed in USA. Â©1997 American Society for Clinical Nutrition

An insulinindexof foods:the insulindemandgeneratedby

1000-kJ portions of common foods13

Susanne HA Holt, Janette C Brand Miller, and Peter Petocz Downloaded from https://academic.oup.com/ajcn/article-abstract/66/5/1264/4655967 by guest on 26 February 2019 [Annotation 1]

#### Page 2

Food Variety, manufacturer, or place of purchase Preparation

INSULIN INDEX OF FOODS 1265

TABLE! Descriptionand preparationof the test foods

Fruit

Black grapes Apples Oranges Bananas Bakery products

Croissants Chocolate cake with

frosting Doughnuts with cinnamon

sugar

Chocolate chip cookies

Water crackers Snack foods and confectionery

Mars Bar

Yogurt

Ice cream

Jellybeans (assorted colors) Peanuts Potato chips

Popcorn

Protein-rich foods

Cheese

Eggs

Lentils

Baked beans

Beefsteak

White fish

Carbohydrate-rich foods

White bread

Whole-meal bread

Grain bread

White rice

Brown rice

White pasta Brown pasta

Potatoes

Waltham cross Reddelicious Navel Cavendish

Fresh, stem removed, served whole Fresh, unpeeled, cut into eight segments Fresh, peeled, cut into eight segments Fresh, peeled, cut into quarters

Defrosted, reheated at 180Â°Cfor 6 mm, and served warm Prepared according to manufacturer's directions, stored at

4 Â°Cup to 2 d before serving at room temperature Prepared by supermarket from standard recipe, defrosted

overnight, reheated at 180 Â°Cfor 5 mm, and served warm Served crisp at room temperature, stored in airtight

container Served crisp at room temperature

Cut into four standard pieces and served at room

temperature Stored at 4 Â°C,served cold

Stored frozen and served cold

Served at room temperature, stored in airtight container Served at room temperature, stored in airtight container Served from freshly opened packet

Prepared according to manufacturer's directions

immediately before serving

All servings cut from same large block, stored at 4 Â°C,

served cold Poached the day before serving, stored at 4 Â°Covernight,

reheatedin microwaveovenfor 1.5mmimmediately before serving Prepared in bulk according to recipe, stored at 4 Â°Cfor up

to 2 d, reheated in a microwave oven for 2 mm immediatelybeforeserving Heated on stove for 5 mm immediately before serving

Grilled the day before serving, cut into standard bite-sized

pieces, and stored at 4 Â°Covernight; reheated in microwave oven for 2 mm immediately before serving Steamed the day before serving, stored at 4 Â°Covernight,

cut into bite-sized pieces, and reheated in microwave oven for 2 miii immediately before serving

Served fresh and plain at room temperature

Served fresh and plain at room temperature

Served fresh and plain at room temperature

Boiled 12 miii and stored overnight at 4 Â°C,reheated in

microwave oven for 1.5 mm immediately before serving Boiled 12 mm and stored overnight at 4 Â°C,reheated in

microwave oven for 1.5 mm immediately before serving Boiled8 mmand storedovernightat 4 Â°C Reheated in microwave oven for 1.5 mm immediately

before serving Peeled,boiledfor 20 mm,and storedat 4 Â°Covernight;

reheated in a microwave oven for 2 mm immediately before serving

Purchased in bulk from supermarket and stored frozen White Wings Foods, Smithfield, Sydney, Australia

Purchased in bulk from supermarket and stored frozen

Anion's Biscuits Ltd. Homebush, Sydney, Australia

Grocery Wholesalers Ltd, Yennora, Australia

Mars Confectionary Australia, Ballarat, Australia

Strawberry fruit yogurt; Australian Co-operative

Foods,' Wetherill Park, Sydney, Australia Vanilla ice cream; Dairy Bell, Camperdown, Sydney,

Australia Grocery Wholesalers Ltd Salted roasted peanuts; Grocery Wholesalers Ltd Crinkle cut chips; Smith's Snackfood Company,

Chatswood, Sydney, Australia Microwave cooked popcorn; Uncle Toby's Company

Ltd, Wahgunyah, Australia

Mature cheddar cheese; Grocery Wholesalers Ltd

Poached hens eggs

Served in tomato sauce2

Canned navy beans in tomato sauce; Franklins,

Chullora, Sydney, Australia Lean topside beef fillets bought in bulk from

supermarket, trimmed and stored frozen

Ling fish ifilets bought in bulk from Sydney fish

markets, trimmed and stored frozen

Fresh sliced wheat-flour bread; Quality Bakers

Australia Ltd. Eastwood, Sydney, Australia Fresh sliced bread made from whole-meal wheat flour;

Riga Bakeries, Moorebank, Sydney, Australia Fresh sliced rye bread containing 47% kibbled rye; Tip

TopBakeries,Chatswood,Sydney,Australia Cairose rice (Sunwhite), Ricegrowers' Co-operative

Ltd. Leeton, Australia Calrose rice (Sunbrown), Ricegrowers' Co-operative

Ltd Spirals Whole-meal spirals; San Remo Pasta Company,

Auburn, Sydney, Australia Russet potatoes

Downloaded from https://academic.oup.com/ajcn/article-abstract/66/5/1264/4655967 by guest on 26 February 2019

#### Page 3

FoodVariety, manufacturer, or place ofpurchasePreparationFrench

friesPrefried oven-baked French fries; McCain's Foods

(Australia), Castle Hill, Sydney, AustraliaStored

frozen, cooked in conventional oven for 15 mm immediately beforeservingBreakfast cereals3CornflakesKellogg's

Australia Pty Ltd, Pagewood, Sydney, Australiaâ€”Special

KToasted flakes made from wheat and rice flour, high in protein; Kellogg's Australia PlyLtdâ€”HoneysmacksPuffed

whole-wheat grains with a honey-based coating; Kellogg's Australia PlyLtdâ€”SustainA mixture of wheat, corn, and rice flakes; rolled oats; dried fruit; and flaked almonds; Kellogg's Australia PtyLtdâ€”All-BranA high-fiber cereal made from wheat bran; Kellogg's Australia PtyLtdâ€”Natural muesliBased on raw rolled oats, wheat bran, dried fruit, nuts,

and sunflower seeds; Uncle Toby's Company Ltd. Wahgunyah,Australiaâ€”PorridgeUncle

Toby's Company Ltd. Wahgunyah, AustraliaRaw rolled oats cooked in a microwave oven according to manufacturer's directions and served without sweetener

1266 HOLT ET AL

TABLE 1 Continued

1 Now Dairy Farmer's. 2 Recipe: 15 mL olive oil, 350 g dried green lentils, 410 g canned tomatoes, 120 g onion, 1 clove garlic, and 1 tsp pepper. 3 All cereals were served fresh with 125 mL fat-reduced (1.5% fat) milk.

of foods.However,carbohydrateis not the only stimulusfor insulin secretion. Protein-rich foods or the addition of protein to a carbohydrate-rich meal can stimulate a modest rise in insulin secretion without increasing blood glucose concentra tions, particularly in subjects with diabetes (20â€”22).Similarly, adding a large amount of fat to a carbohydrate-rich meal increases insulin secretion even though plasma glucose re sponses are reduced (23, 24).

Thus, postprandial insulin responses are not always propor tional to blood glucose concentrations or to a meal's total carbohydrate content. Several insulinotropic factors are known to potentiate the stimulatory effect of glucose and mediate postprandial insulin secretion. These include fructose, certain amino acids and fatty acids, and gastrointestinal hormones such as gastric inhibitory peptide, glucagon, and cholecystokiin [Annotation 6] (25, 26). Thus, protein- and fat-rich foods may induce substan tial insulin secretion despite producing relatively small blood glucose responses. We therefore decided that comparing the insulinemic effects of foods on an isoenergetic basis was a logical and practical approach.

The aim of this study was to systematically compare post prandial insulin responses to isoenergetic portions of a range of common foods. An insulin score (IS) was calculated for each food on the basis of its insulinemic effect relative to a reference food. Thirty-eight foods, categorized into six different food groups, were studied to determine which foods within the same food group were most insulinogenic. We hypothesized that postprandial insulin responses are not closely related to the carbohydrate content or glycemic effects of some foods.

SUBJECTS AND METHODS

Test foods

Thirty-eight foods were tested and were grouped into six food categories: 1) fruit: grapes, bananas, apples, and oranges;

2) bakery products: croissants, chocolate cake with icing, doughnuts with cinnamon sugar, chocolate chip cookies, and water crackers; 3) snack foods and confectionery: Mars Bar candy bar (Mars Confectionary Australia, Ballarat, Australia), strawberry yogurt, vanilla ice cream, jellybeans, salted roasted peanuts, plain potato chips, and plain popcorn; 4) protein-rich foods: cheddar cheese, poached eggs, boiled lentils in a tomato sauce, baked beans in a tomato sauce, grilled beef steak, and steamed white fish; 5) carbohydrate-rich foods: white bread, whole-meal bread, rye-grain bread, white rice, brown rice, white pasta, brown pasta, boiled potatoes, and oven-baked French fries; and 6) breakfast cereals: Cornflakes (Kellogg's Australia Pty Ltd. Pagewood, Australia), Special K (Kellogg's Australia Pty Ltd), Honeysmacks (Kellogg's Australia Pty Ltd), Sustain (Kellogg's Australia Pty Ltd), All-Bran (Kellogg's Australia Pty Ltd), natural muesli, and oatmeal porridge.

Each food was served plain as a 1000-U portion with 220 mL water. White bread was used as the reference food for each food group. The foods were selected to represent a range of natural and processed foods commonly eaten in industrialized societies. Details of the foods and their preparation methods are listed in Table 1. Foods were bought in bulk to minimize variations in composition and were served in standard-sized pieces. The nutritional composition ofeach food per 1000 U as calculated from Australian food tables or manufacturers' data is shown in Table 2.

Subjects

Separate groups of healthy subjects (n = 11â€”13)were re cruited to test each category of foods. Volunteers were ex cluded if they were smokers or taking prescription medications, had a family history of diabetes or obesity, were dieting, or had irregular eating habits. In total, 41 subjects participated. One subject consumed all of the test foods and 15 other subjects

Downloaded from https://academic.oup.com/ajcn/article-abstract/66/5/1264/4655967 by guest on 26 February 2019

#### Page 4

ServingCarbohydrateEnergyFood

weight FatProtein SugarStarch FiberWater density

INSULIN INDEX OF FOODS 1267

TABLE 2 Nuthtional composition of the test foods per 1000-U serving as calculated from Australian food tables or manufacturers' data'

g g g g g g k@I/gFruitGrapes3950.43.256.90.03.6317.02.5Bananas2790.34.747.28.46.1210.13.6Apples4350.01.356.52.29.1360.92.3Oranges6250.66.950.60.012.5539.41.6Bakery

productsCroissants6114.46.13.118.61.813.516.4Cake26411.94.320.110.50.710.715.6Doughnuts6513.44.38.917.01.416.115.4Cookies25110.92.418.716.21.02.119.6Crackers585.45.81.340.21.62.217.2Snacks

andconfectioneryMars Barr549.42.936.71.11.73.518.5Yogurt@2415.311.837.60.00.5187.04.2Ice

cream12013.45.225.80.00.074.28.3Jellybeans880.05.344.611.50.012.211.4Peanuts3820.19.61.73.72.40.626.3Potatochips24416.22.70.222.12.41.122.7Popcorn24713.04.62.125.36.21.721.3Protein-rich

foodsCheese5920.015.00.10.00.020.916.9Eggs15917.919.60.50.00.0119.46.3Lentils2534.619.44.224.91

1.4222.03.9Baked beans3511.716.116.123.216.8267.12.8Beef steak1587.742.00.00.00.0104.36.3Fish3331.056.30.00.00.0250.03.0Carbohydrate-rich

foodsWhite bread2942.18.51.844.13.336.110.6Whole-meal bread21012.67.61.743.76.640.39.9Grain bread21085.49.42.437.66.541.49.3White rice22030.55.00.156.00.4140.04.9Brown rice21482.15.20.552.61.493.96.8White pasta2010.87.82.047.13.5134.85.0Brown pasta22181.611.30.747.810.9132.64.6Potatoes3681.010.03.145.99.2290.82.7French

fries2938.73.91.135.43.533.810.7Breakfast

cerealsCornflakes21702.18.410.236.11.5110.95.9Special

K21722.115.314.027.21.4111.25.8Honeysmacks21722.28.731.117.02.61 15.05.8Sustain21683.19.713.729.13.2119.15.9Muesli21756.110.717.119.86.61

14.15.7Pomdge23836.210.97.529.04.7333.72.6All-Bran21742.911.713.929.414.1111.05.7

I Mars Bar, Mars Confectionary Australia, Ballarat, Australia; Comfiakes, Special K, Honeysmacks, Sustain, and All-Bran: Kellogg's Australia Pty Ltd. Pagewood, Australia.

2 Nutrient composition calculated from manufacturer's data.

completed two or more food categories. All of the subjects approved by the Medical Ethical Review Committee of the were university students; relevant characteristics ofthe subjects University of Sydney. are listed in Table 3. The mean body mass index (BMI, in kg/m2) of the 41 subjects was 22.7 Â±0.4 (range: 19â€”29).Three P1@Ot(WOl subjects had a BMI > 25 but two of these were short, stocky Each subject first consumed a 1000-U portion of white bread males who had excess muscle rather than fat. Female subjects (45.9 g carbohydrate) to confirm normal glucose tolerance. did not participate during their menstrual period or if they White bread was also used as the reference food (IS = 100%) [Annotation 7] experienced adverse premenstrual symptoms. Informed con- against which all other foods were compared, similar to the sent was obtained from all of the subjects and the study was method used for calculating GI values of foods (18). The use of [Annotation 8]

Downloaded from https://academic.oup.com/ajcn/article-abstract/66/5/1264/4655967 by guest on 26 February 2019

#### Page 5

FoodgroupAgeBMI2yFruit

(n = 5 F, 6 M)22.9 Â±3.922.9 Â±1.4Bakery products (n = 6 F, 6 M)22.2 Â±3.723.1 Â±2.7Snacks andconfectionery(n = 5 F, 7 M)21.0 Â±1.222.9 Â±3.5Protein-rich foods (n 5 F, 6 M)22.4 Â±2.824.3 Â±3.1Carbohydrate-rich foods (n = 5 F, 8 M)21.0 Â±1.923.0 Â±1.9Breakfast cereals (n = 5 F, 6 M)22.8 Â±3.922.8 Â±1.4

1268 HOLT ET AL

tube radioimmunoassay kit (Coat-A-Count; Diagnostic Prod ucts Corporation, Los Angeles). For both plasma glucose and insulin analysis, all nine plasma samples for a particular sub ject's test were analyzed within the same run to reduce any error introducedby interassayvariation.When possible,all plasma samples for a particular subject were analyzed for insulin within the same run. For the insulin analysis, the mean within-assay CV was 5% and the mean between-assay CV was 7%.

Statistical analysis

Cumulative changes in postprandial plasma glucose and insulin responses for each food were quantified as the incre mental area under the 120-mn response curve (AUC), which was calculated by using the trapezoidal rule with fasting con [Annotation 9] centrations as the baseline and truncated at zero. Any negative areas tended to be small and were ignored. For each subject, an IS (%) was calculated for each test food by dividing the insulin AUC value for the test food by the insulin AUC value for white bread (the reference food), and expressed as a percentage as follows:

IS (%)

Area under the 120-mm insulin response

curve for 1000 U test food Areaunderthe 120-mminsulinresponsecurve

for 1000 Id white bread

TABLE 3 Characteristics of each group of subjects'

â€˜¿I Â± SD. 2 In kg/rn2.

a reference food controls for inherent differences between individuals that affect insulin sensitivity, such as body weight and activity levels.

Subjects were fed 1000-U portions of the test foods in a random order on separate mornings after a 10-h overnight fast. Within each food group, each subject acted as his or her own control, being tested at the same time of day and under [Annotation 10] as similar conditions as possible. Subjects were asked to refrain from unusual activity and food intake patterns, to abstain from alcohol and legumes the day before a test, and to eat a similar meal the night before each test. When subjects arrived at the lab in the morning, they completed a short questionnaire assessing recent food intake and activity patterns. A fasting finger-prick blood sample was collected and subjects were then given a test food and 220 mL water (0 mm). When possible, foods were presented under a large opaque plastic hood with a hole through which volunteers pulled out pieces of the test food one at a time. This was an attempt to minimize between-subject variation in cephalic phase insulin secretion arising from the sensory stimulation associated with the anticipation and act of eating (27). However, this was not feasible for the liquid foods (yogurt and ice cream), foods served in a sauce (baked beans and lentils), or with milk (all of the breakfast cereals), which were presented in standard bowls without the hood.

Subjects were asked to eat and drink at a comfortable rate. Immediately after finishing the test food, subjects recorded the time taken to eat the food and completed a questionnaire assessing various appetite responses and the food's palatability. [These results are reported in a separate paper (28).] Subjects remained seated at tables in a quiet environment and were not permitted to eat or drink until the end of the session (120 mm).

Finger-prick blood samples (1.5â€”2.5mL) were collected from warmed hands immediately before the meal (0 mm) and [Annotation 11] 15, 30, 45, 60, 75, 90, 105, and 120 mm after the start of the meal (into plastic tubes that had been kept on ice) with use of an automatic lancet device (Autoclix; Boehringer Mannheim Australia, Castle Hill, Australia). Blood samples were centri fuged immediately after collection (1 miii at 12 500 X g at room temperature) and plasma was pipetted into chilled tubes and immediately stored at â€”¿20Â°Cuntil analyzed (< 1 mo). Plasma glucose concentrations were analyzed in duplicate with a Cobas Fara automatic spectrophotometric analyzer (Roche Diagnostica, Basel, Switzerland) and the glucose hexokinase enzymatic assay. The mean within-assay and between-assay precisions (CVs) were both < 6%. Plasma insulin concentra tions were measured in duplicate by using an antibody-coated

X100 (1)

This equation is similar to that developed by Wolever and Jenkins (29) for calculating GI values. A glucose score (GS) (not the same as a GI score, which is based on a 50-g carbo hydrate portion) for each food was also calculated by using the same equation with the corresponding plasma glucose results.

Analysis of variance (ANOVA) and Fisher's probable least significant-difference test for multiple comparisons were used to determine statistical differences among the foods within each food group (STATVIEW STUDENT SOFFWARE; Abacus Concepts mc, Berkley, CA). Linear-regressionanalysis was used to test associations between glucose and insulin responses and nutritional indexes (MINITAB DATA ANALYSIS SOFFWARE, version 7.0; Minitab Inc, State College, PA). Test foods not containing a particular nutrient were excluded from these analyses. Therefore, sample sizes for the correlations between individual nutrients and the mean GSs and ISs varied from 32 to 36. Mean results for white bread for each food group were included in some statistical anal yses, so these correlations were made with 43 values. One subject from the protein-rich food group did not complete the fish test and one subject from the breakfast cereal group did not complete the Sustain test. Therefore, in total, 503 indiVidUal tests were fully completed.

Stepwise-multiple-regression analysis was used to examine the extent to which the different macronutrients and GSs ac counted for the variability of the ISs (MINITAB DATA ANALYSIS SOFTWARE). For this analysis, the individual white bread OS and IS results were included for the carbohy drate-rich food group only; therefore, this analysis was per formed with 446 individual observations for 38 foods. Includ ing the white bread results for each food group (n = 503) suggests that independent repeat tests were done for white

Downloaded from https://academic.oup.com/ajcn/article-abstract/66/5/1264/4655967 by guest on 26 February 2019

#### Page 6

Food Glucose AUCInsulin AUCInsulin AUC:

glucose AUCInsulin AUC per g carbohydrateInsulin AUC per g serving weightGlucose scoreInsulinscoremol-miniLpmolmin/LpmolminL'g'pmolminL'g'%%

INSULIN INDEX OF FOODS 1269 TABLE 4 Areas under the 120-mm plasma glucose and insulin response curves (AUCs), ratio of insulin AUC to glucose AUC, the insulin AUC per g carbohydrate and per g serving weight, and mean glucose and insulin scores'

BreakfastcerealsWhitebread156Â±2113557Â±1756108Â±19295Â±38144Â±19100Â±0100Â±0All-Bran59

Â±94299 Â±61287 Â±1599 Â±1425 Â±340 Â±732 Â±4Porridge80 Â±95093 Â±49374 Â±11139 Â±1313 Â±I60 Â±1240 Â±4Muesli65Â±126034Â±813118Â±18163Â±2234Â±543Â±746Â±5Special K106 Â±148038 Â±63595 Â±14195 Â±1547 Â±470 Â±966 Â±5Honeysmacks91Â±109102Â±1506108Â±12189Â±3153Â±960Â±767Â±6Sustain93Â±88938Â±757102Â±9209Â±1853Â±466Â±671Â±6Cornflakes1

10 Â±118768 Â±62388 Â±5189 Â±1352 Â±476 Â±1175 Â±8Groupmeanâ€”7183Â±35792Â±5169Â±839Â±259Â±357Â±3Carbohydrate-rich foodsWhitebread120Â±1312882Â±1901112Â±15281Â±41137Â±20100Â±0100Â±0Whitepasta50Â±114456Â±453156Â±4891Â±922Â±246Â±1040Â±5Brown

pasta74 Â±74535 Â±57467 Â±1093 Â±1221 Â±368 Â±1040 Â±5Grainbread68Â±96659Â±837106Â±12166Â±2162Â±860Â±1256Â±6Brownrice113Â±136240Â±61658Â±5117Â±1142Â±4104Â±1862Â±11French

fries70 [Annotation 12] Â±117643 Â±713146 Â±29209 Â±1982 Â±871 Â±1674 Â±12Whiterice129Â±168143Â±68369Â±5145Â±1240Â±3110Â±1579Â±12Whole-meal bread106 Â±141 1 203 Â±1420122 Â±20247 Â±311 11 Â±1497 Â±1796 Â±12Potatoes148Â±2413930Â±1467120Â±19284Â±3038Â±4141Â±35121Â±11Groupmeanâ€”8410Â±461106Â±8182Â±1062Â±588Â±674Â±8Protein-rich

foodsWhitebread121Â±1917438Â±3154177Â±35387Â±63185Â±33100Â±0100Â±0Eggs36

Â±114744 Â±1017135 Â±929340 Â±184530 Â±642 Â±1631 Â±6Cheese42 Â±105994 Â±1590268 Â±15364 257 Â±15013106 Â±2755 Â±1845 Â±13Beef18Â±67910Â±21931583Â±939â€”50Â±1421Â±851Â±16Lentils63 Â±179268 Â±2174307 Â±103325 Â±6837 Â±962 Â±2258 Â±12Fish29 Â±149350 Â±2055775 Â±502â€”28 Â±628 Â±1359 Â±18Bakedbeans110Â±1420106Â±3776183Â±44504Â±8757Â±11114Â±18120Â±19Group meanâ€”9983 Â±1032585 Â±6118 607 Â±545653 Â±654 Â±761 Â±7FruitWhite bread171 Â±1915 563 Â±1632105 Â±18339 Â±36166 Â±17100 Â±0100 Â±0Apples83Â±78919Â±910118Â±18152Â±1520Â±250Â±659Â±4Oranges66Â±119345Â±1074166Â±23185Â±2115Â±239Â±760Â±3Bananas133

Â±1212445 Â±1353108 Â±22224 Â±2445 Â±579 Â±1081 Â±5Grapes126Â±1412293Â±1190113Â±19216Â±2131Â±374Â±982Â±6Groupmeanâ€”10751Â±605124Â±10194Â±1128Â±261Â±571Â±3Snacks

andconfectioneryWhitebread159Â±2915592Â±2376104Â±24340Â±52166Â±25100Â±0100Â±0Peanuts20

Â±73047 Â±828214 Â±88564 Â±15380 Â±2212 Â±420 Â±5Popcorn71Â±126537Â±679109Â±32239Â±25139Â±1462Â±1654Â±9Potatochips77Â±158195Â±1577169Â±78367Â±71186Â±3652Â±961Â±14Ice

cream93 Â±1712 348 Â±1867172 Â±38479 Â±72103 Â±1670 Â±1989 Â±13Yogurt88Â±2315611Â±1808167Â±33415Â±4865Â±762Â±15115Â±13MarsBar98Â±1016682Â±1896218Â±65441Â±50309Â±3579Â±13122Â±15Jellybeans161Â±1822860Â±368133Â±27407Â±64260Â±41118Â±18160Â±16Groupmeanâ€”12183Â±994191Â±20416Â±30163Â±1465Â±689Â±7Bakery

productsWhitebread129Â±1517599Â±3058188Â±64383Â±67187Â±33100Â±0100Â±0Doughnuts78Â±1412445Â±2402113Â±21480Â±93191Â±3763Â±1274Â±9Croissants89Â±613097Â±2978483Â±244604Â±137215Â±4974Â±979Â±14Cake61

Â±1114 305 Â±3472178 Â±54467 Â±I13223 Â±5456 Â±1482 Â±12Crackers139 Â±2614 673 Â±2686331 Â±104354 Â±65253 Â±461 18 Â±2487 Â±12Cookies92Â±1215223Â±382200Â±57436Â±110298Â±7574Â±1192Â±15Groupmeanâ€”12681Â±1325261Â±56468Â±47236Â±2477Â±783Â±5

I1@ SEM. Mars Bar, Mars Confectionary Australia, Ballarat, Australia; All-Bran, Special K, Honeysmacks, Sustain, and Cornflakes: Kellogg's Australia Pty Ltd. Pagewood, Australia.

Downloaded from https://academic.oup.com/ajcn/article-abstract/66/5/1264/4655967 by guest on 26 February 2019

#### Page 7

1270 HOLT ET AL

bread, which artificially increases the accuracy of any calcu lation involving white bread.

RESULTS

Fasting glucose and insulin concentrations

Within each food group, the subjects' average fasting plasma glucose and insulin concentrations were not signif icantly different among the foods. Mean fasting plasma glucose concentrations did not vary significantly among the six food groups, whereas mean fasting insulin concentra tions were more variable, ranging from â€œ¿42to 120 pmol/L. Fasting insulin concentrations were not more variable in females than in males and there were no significant differ ences at various stages of the menstrual cycle. A significant correlation was found between mean fasting insulin concen trations and mean BMI values for the six groups of subjects [Annotation 14] (r â€”¿0.81,P < 0.05,n 6).

Postprandial glucoseand insulin responses

As with any biological response, there was between-subject variation in the glucose and insulin responses to the same food. Two-way ANOVA was used to examine the ranking of each subject's responses to the different test foods within a food group (ie, interindividual variation). There were significant differences among the subjects in the rank order of their glu cose AUC responses except within the fruit and protein-rich food groups. There were also significant differences among the subjects' rank order of insulin AUC responses within all food groups. However, individual subjects within each food group consistently produced relatively low, medium, or high insulin responses. Furthermore, subjects produced their lowest insulin responses for the least insulinogenic foods and their highest insulin responses for the most insulinogenic foods within each food group.

100

There were large differences in mean glycemic and insulin responses to the foods, both within and between food groups. [Annotation 13] Mean glucose and insulin AUC results, mean GSs and ISs, and the mean ratios of insulin to glucose AUCs (the amount of insulin secretion in relation to the blood glucose response) are listed in Table 4. Mean GSs and ISs were calculated for each food group by averaging the scores for all test foods within the food group. On average, the snack food group produced the highest food group IS (89%), followed by bakery products (83%), carbohydrate-rich foods (74%), fruit (71%), protein rich foods (61%), and breakfast cereals (57%). Average GSs for the food groups did not follow the same rank order (Figure 1). The carbohydrate-rich food group produced the highest average GS (88%), followed by bakery products (77%), snack foods (65%), fruit (6 1%), breakfast cereals (59%), and protein rich foods (54%). Interestingly, the GS rank order is not pro portional to the average total carbohydrate content of each food group, which highlights the influence of other food factors (eg, fiber and processing) in determining the rate of carbohydrate digestion and absorption.

Overall, among the 38 test foods, jellybeans produced the [Annotation 15] highest mean IS (160 Â±16%), eightfold higher than the lowest IS (for peanuts: 20 Â±5%) (Figure 2). White bread, the stan dard food, consistently produced one of the highest glucose and insulin responses (peak and AUC) and had a higher IS than most of the other foods (84%). All of the breakfast cereals were significantly less insulinogenic than white bread (P < 0.001). All-Bran and porridge both produced a significantly lower IS than the other cereals (P < 0.001), except muesli. Despite containing more carbohydrate than porridge and muesli, All Bran produced the lowest GS. Baked beans, which contain considerably more carbohydrate than the other protein-rich foods, produced a significantly higher GS and IS (P < 0.001). On average, fish elicited twice as much insulin secretion as did the equivalent portion of eggs. Within the fruit group, oranges and apples produced a significantly lower GS and IS than

0 Glucosescore . Insulinscore

80

60J

40

20

0

4) 0 U â€˜¿I) C 4) 4) E 0. 0 0

Breakfast cereals

Carbohydrate- Bakery

rich foods products

Protein rich foods

Fruit Snacksand

confectionery

FIGURE 1. Mean(Â±SEM)glucoseand insulinscoresfor each foodgroup.

Downloaded from https://academic.oup.com/ajcn/article-abstract/66/5/1264/4655967 by guest on 26 February 2019

#### Page 8

...@White

breac.IPrthifr@c.

Doughnuts Croissants

Cake Crackers Cookies

PeanutsI=@lIIIâ€•;â€•t-1@PopcornI=â€•iiPotato

chipsI='=â€•iIcecreamlâ€•;lIYogurtI

. lip

INSULIN INDEX OF FOODS 1271

I@1

ft

All-Bran Porridge

Muesli

OUSL4I(J

Eg@ Cheese

Beef Lentils

Fish Beans

Apples Oranges Bananas Grapes

Brown pasta White pasta Grain bread Brownrice Frenchfries

White rice Whole-meal bread

-i .1

1=1

Jellybeans

100 200

Insulin score (%)

FIGURE 2. Mean (Â±SEM) insulin scores for 1000-Id portions of the test foods. White bread was the reference food (insulin score = 100%).All-Bran cereal, Special K cereal, Honeysmacks cereal, Sustain cereal, and Cornflakes, Kellogg's Australia Pty Ltd. Pagewood, Australia; Mars Bar candy bar, Mars Confectionary Australia, Ballarat, Australia.

grapes and bananas (P < 0.05 to P < 0.001), despite contain ing a similar amount of carbohydrate.

Potatoes produced significantly higher GSs and ISs than all of the other carbohydrate-rich foods. White bread produced a higher GS and IS than grain bread (P < 0.05 and P < 0.001 respectively), but whole-meal bread and white bread had sim ilar scores. White and brown rice had similar GSs and ISs, as did white and brown pasta. Among the bakery products, crack ers produced a significantly higher GS than the other test foods, but there were no significant differences in ISs within this group (all tended to be high). Among the snack foods, jelly beans produced a significantly higher GS and IS than the other foods in this group. Despite containing similar amounts of carbohydrate, jellybeans induced twice as much insulin secre tion as any of the four fruits. The candy bar and yogurt, which

both contained large amounts of sugar in combination with fat or protein, produced relatively high ISs. Popcorn and potato chips elicited twice as much insulin secretion as peanuts (P < 0.05 and P < 0.01, respectively).

Significant differences were found both within and among the food groups when the insulin AUC responses were examined as a function of the food's carbohydrate content (Table 4). On average, protein-rich foods produced the highest insulin secretion per gram of carbohydrate (food group mean: 18 607 pmol . mm . L' . g@1) (because of their mostly low carbohydrate contents), followed by bakery products (468 pmol . mm . L@ . g1), snack foods (416 pmol . mm . L â€˜¿. g 1) fruit (194 pmol . mm . L â€˜¿. g 1), carbohydrate-rich foods (182 pmol . mm . L@ . g'), and breakfast cereals ( 169 pmol . mm . L â€˜¿. g â€˜¿).When the

Downloaded from https://academic.oup.com/ajcn/article-abstract/66/5/1264/4655967 by guest on 26 February 2019

#### Page 9

1272 HOLT ET AL

S

insulin AUC response was examined in relation to the food's serving size (g), the bakery products were the most insuli nogenic (food group mean: 236 pmol . mm . L â€˜¿. g 1), fol lowed by snack foods (163 pmol . mm . L â€˜¿. g 1), carbo hydrate-rich foods (62 pmol . mm . L â€˜¿. g 1), protein-rich foods (53 pmol . mm . L â€˜¿. g 1), breakfast cereals (39 pmol . mm . L â€˜¿. g â€˜¿),and fruit (28 pmol . mm . L â€˜¿. g 1). These results reflect the insulinogenic effects of protein and fat.

Insulin responses in relation to glucose responses

Overall, mean glucose and insulin AUC values were posi tively correlated (r = 0.67, P < 0.001, n = 43), as were the peak glucose and insulin values (r = 0.57, P < 0.001, n = 43). Hence, the mean GSs and ISs were highly correlated (r = 0.70, P < 0.001, n = 38) (Figure 3). The peak glucose concentration (change from fasting) correlated positively with glucose AUC values (r = 0.74, P < 0.001, n = 43) and peak insulin concentrations were proportional to the insulin AUC values (r 0.95, P < 0.001, n 43). In addition, the observed GSs for 1000-U portions of the foods correlated with previously published GI values based on portions of foods containing 50 g carbohydrate (r = 0.65, P < 0.001, n = 32). Six test foods (chocolate chip cookies, eggs, cheese, beef, fish, and Hon eysmackscereal)werenotincludedin thisanalysisbecauseGI values were not available.

Insulin AUC values were divided by glucose AUC values to determine which foods were markedly insulinogenic relative to their glycemic effect (Table 4 and Figure 4). On average, the protein-richfoodsstimulateda largeamountof insulinsecre tion relative to their glycemic response, followed by the bakery products, snack foods, fruit, carbohydrate-rich foods, and breakfast cereals.

4) 1@ 0 U 4) C 4) C

200

FIGURE3. Relationbetweenthemeanglucoseandinsulinscores(r = 0.74, P < 0.001, n = 38).

Relationsbetweenmetabolic responsesand nutrient contents of the foods

Correlations between the macronutrient compositions of the test foods and the mean ISs are shown in Figure 5. The portion size (energy density: kJ/g), water, and fiber contents of the foods were not significantly related to the mean ISs. The relation between protein contents and ISs was negative but not significant (r â€”¿0.24,n = 38). The mean ISs were positively related to total carbohydrate content (r = 0.39, P < 0.05, n = 36) and sugar content (r = 0.36, P < 0.05, n = 36), but were not significantly related to starch content (r = â€”¿0.09,n = 30). Fat content was negatively related to the mean IS (r = â€”¿0.27, NS, n = 36). When expressed as a percentage of total energy, fat (r = â€”¿0.27,NS, n = 36) and protein (r = â€”¿0.24,NS, n = 38) were negatively associated with the mean IS, whereas total carbohydrate was positively related (r = 0.37, P < 0.05, n 36).

Relations between the GSs and the nutrients largely followed the same directions as the IS correlations. Mean GSs were not significantly related to the foods' serving sizes or water or fiber contents. Mean GSs correlated negatively with fat (r = â€”¿0.38, P < 0.05, n = 36) and protein (r = â€”¿0.38,P < 0.05, n = 38) contents, and positively with total carbohydrate content (r = 0.32, NS, n = 36). Unlike the ISs, the GSs were significantly related to starch content (r = 0.43, P < 0.05, n = 30) but not sugar content (r = â€”¿0.07,NS, n = 36). When expressed as a percentageof total energy, fat (r = â€”¿0.38,P < 0.05, n = 36) and protein (r = â€”¿0.39,P < 0.05, n = 38) were negatively associated with mean GSs, whereas total carbohydrate content was positively related (r = 0.46, P < 0.01, n = 36).

Stepwise-multiple-regression analysis of the 446 individual results for the 38 foods was performed to determine the extent to which the macronutrients and GSs accounted for the van ability of the ISs. Unfortunately, it was not possible to generate a single multiple-regression equation that included all of the macronutrients because some pairs of nutrients were highly correlated (eg, fat and protein, fiber and water, total carbohy drate and sugar or starch, and sugar and starch). The regression equation that included all of the macronutrients had unaccept ably high variance inflation factors. Therefore, two separate regression equations were generated that were limited to the factors that were measured and not interdependent. Equation 2 includes fat but not protein, whereas equation 3 includes pro tein but not fat:

IS = 72.4 + 0.383 GS â€”¿1.88 fat â€”¿0.103 water

+ 0.509 sugar â€”¿0.421 starch (2)

for whichSD = 37.34,R2= 33.1%,andadjustedR2= 32.4%. P values (significance found in the linear-regression analysis for the associations between the individual nutrients and the IS) are as follows: OS and water (P < 0.000), fat (P < 0.001), sugar (P < 0.005), and starch (P < 0.036).

IS = 23.2 + 0.383 05 + 0.785 protein â€”¿0.098 water

+ 1.29 sugar + 0.377 starch (3) [Annotation 16]

for which SD = 37.42, R2 = 32.8%, and adjusted R2 = 32.1%. P values are as follows: GS, water, and sugar (P < 0.000); protein (P < 0.003); and starch (P < 0.02).

.

S

â€¢¿â€¢

S

0 0 100

Glucose score (%) [Annotation 17]

Downloaded from https://academic.oup.com/ajcn/article-abstract/66/5/1264/4655967 by guest on 26 February 2019

#### Page 10

INSULIN INDEX OF FOODS 1273

Porridge All-Bran Comfiakes Special K

Sustain Honeysmacks

Muesli

Brownrice Brownpasta

White rice Grain bread Whitebread

Potatoes Whole-meal bread

Frenchfries White pasta

Bananas Grapes Apples Oranges

Popcorn Jellybeans

Yogurt Chips Icecream Peanuts MarsBar

Doughnuts

Cake Cooides Crackers Croissant

Eggs Beans Cheese

Lentils

Fish Be@

I.. I-,

I-i

@1 -I

-t

0 500 1000 1500 2000 2500 3000

InsulinAUC/GiucoseAUC

FIGURE 4. Ratio of insulin area under the curve (AUC) to glucose AUC responses. I Â±SEM. All-Bran cereal, Special K cereal, Honeysmacks cereal, Sustaincereal,andCornflakes,Kellogg'sAustraliaPtyLtd.Pagewood,Australia;MarsBarcandybar,MarsConfectionaryAustralia,Ballarat,Australia.

Linear-regression analysis of the individual OS and IS re sults had an R2 value of 23%. Therefore, the glycemic response was a significant predictor of the insulin response, but it accounted for only 23% of the variability in insulinemia. The macronutrients (protein or fat, water, sugar, and starch) were also significant predictors but together accounted for only another 10% of the variability of the insulin responses. Thus, we can explain only 33% of the variation of the insulin re sponses to the 38 foods studied.

DISCUSSION

The results of this study confirm and also challenge some of our basic assumptions about the relation between food intake

and insulinemia. Within each food group, there was a wide range of insulin responses, despite similarities in nutrient corn position. The important Western staples, bread and potato, were among the most insulinogenic foods. Similarly, the highly refined bakery products and snack foods induced substantially more insulin secretion per kilojoule or per gram of food than did the other test foods. In contrast, pasta, oatmeal porridge, and All-Bran cereal produced relatively low insulin responses, despite their high carbohydrate contents. Carbohydrate was quantitatively the major macronutrient for most foods. Thus, it is not surprising that we observed a strong correlation between GSs and ISs (r = 0.70, P < 0.001). However, some protein and fat-rich foods (eggs, beef, fish, lentils, cheese, cake, and doughnuts) induced as much insulin secretion as did some

Downloaded from https://academic.oup.com/ajcn/article-abstract/66/5/1264/4655967 by guest on 26 February 2019

#### Page 11

0

C U) C

0 20 40 60

Protein (glservlng)

100

S S

0 20 40 60

1274 HOLT ET AL

200

100

0

140

S S 5

S S

S

S

S.

â€¢¿t,.. S

SIâ€¢ S

0 10 20

Fiber (9/serving)

200

100

S S

S U) C a, C

S U) C U) C

S

@fS55 I

Starch (9/serving)

S S

200W

100

0

@S S

S 55

55 S S

1@b5 s

S

Fat (9/serving)

S

0 20 40 60

Total carbohydrate (9/serving)

0 i I

0 20 40 60

Sugar (9/serving)

140

S

100

S

S S 0 S S

S

S S

60

7c1 0

0 10 20 30

FIGURE 5. Relations between the nutrient contents of the test foods and the mean insulin scores. Fiber: r = â€”¿0.10,NS, n = 32; protein: r = â€”¿0.24, [Annotation 18] NS,n = 38; totalcarbohydrate:r = 0.39,P < 0.05,n = 36; sugar:r = 0.36,P < 0.05,n 36; starch:r = â€”¿0.09,NS,n = 30; fat: r = â€”¿0.27,P < 0.05, n = 36.

carbohydrate-rich foods (eg, beef was equal to brown rice and fish was equal to grain bread). As hypothesized, several foods with similar GSs had disparate ISs (eg, ice cream and yogurt, brown rice and baked beans, cake and apples, and doughnuts and brown pasta). Overall, the fiber content did not predict the magnitude of the insulin response. Similar ISs were observed for white and brown pasta, white and brown rice, and white and

whole-meal bread. All of these foods are relatively refined compared with their traditional counterparts. Collectively, the fmdings imply that typical Western diets are likely to be significantly more insulinogenic than more traditional diets based on less refined foods.

In this study, we chose to test isoenergetic portions of foods rather than equal-carbohydrate servings to determine the insu

Downloaded from https://academic.oup.com/ajcn/article-abstract/66/5/1264/4655967 by guest on 26 February 2019

#### Page 12

INSULIN INDEX OF FOODS 1275

lin response to all of the nutrients in the foods as normally consumed. A standard portion size of 1000 kJ was chosen because this resulted in realistic serving sizes for most of the foods except apples, oranges, fish, and potatoes. Although some of the protein-rich foods may normally be eaten in smaller quantities, fish, beef, cheese, and eggs still had larger insulin responses per gram than did many of the foods consist ing predominantly of carbohydrate. As observed in previous studies, consumption of protein or fat with carbohydrate in creases insulin secretion compared with the insulinogenic ef fect of these nutrients alone (22, 30â€”32). This may partly explain the markedly high insulin response to baked beans. Dried hancot beans, which are soaked and boiled, are likely to have a lower IS than commercial baked beans, which are more readily digestible.

The results confirm that increased insulin secretion does not account for the low glycemic responses produced by low-GI foods such as pasta, porridge, and All-Bran cereal (33). Fur thermore, equal-carbohydrate servings of foods do not neces sarily stimulate insulin secretion to the same extent. For exam ple, isoenergetic servings of pasta and potatoes both contained =,%50g carbohydrate, yet the IS for potatoes was three times greater than that for pasta. Similarly, porridge and yogurt, and whole-grain bread and baked beans, produced disparate ISs despite their similar carbohydrate contents. These findings, like others, challenge the scientific basis of carbohydrate exchange tables, which assume that portions of different foods containing 10â€”15g carbohydrate will have equal physiologic effects and will require equal amounts of exogenous insulin to be metab olized. It is possible that preprandial insulin doses for patients with NIDDM could be more scientifically estimated or matched on the basis of a meal's average insulinemic effect in healthy individuals, rather than on the basis of the meal's carbohydrate content or 01. Further research is required to test this hypothesis. The advent of intensive insulin therapy and the added risk of hypoglycemia increases the urgency of this research (34).

Our study was undertaken to test the hypothesis that the postprandial insulin response was not necessarily proportional to the blood glucose response and that nutrients other than carbohydrate influence the overall level of insulinemia. Multi pIe-regression analysis of the individual results showed that the glycemic response was a significant predictor of the insulin response, but it accounted for only 23% of the variability in [Annotation 19] insulinemia. The macronutrients (protein or fat, water, sugar, and starch) were also significant predictors, but together ac counted for only another 10% of the variability of the insulin responses. Thus, we can explain only 33% of the variation of [Annotation 20] the insulin responses to the 38 foods under examination. The low R2 value indicates that the macronutrient composition of foods has relatively limited power for predicting the extent of postprandial insulinemia. The rate of starch digestion, the amount of rapidly available glucose and resistant starch, the degree ofosmolality, the viscosity ofthe gut's contents, and the rate of gastric emptying must be other important factors influ encing the degree of postprandial insulin secretion. Further research is required to examine the relation between postpran dial insulinemia, food form, and various digestive factors for a much larger range of foods to produce a regression equation with greater predictive value.

Dietary guidelines for healthy people and persons with NIDDM have undergone considerable change and will con tinue to be modified as our understanding of the relations between dietary patterns and disease improves. There is con cern that high-carbohydrate diets may increase triacylglycerol concentrations and reduce high-density lipoprotein concentra tions (35, 36). The use of diets high in monounsaturated fat is an attempt to overcome the undesirable effects of some high carbohydrate diets on plasma lipids (37â€”39).However, diets high in monounsaturated fat are unlikely to facilitate weight loss. A low-fat diet based on less-refined, carbohydrate-rich foods with relatively low ISs may help enhance satiety and aid weight loss as well as improve blood glucose and lipid control (4).

The results of this study are preliminary but we hope they stimulate discussion and further research. Additional studies are needed to determine whether the IS concept is useful, reproducible around the world, predictable in a mixed-meal context, and cliii ically useful in the treatment of diabetes mellitus, hyperlipidemia, and overweight. Studies examining the relation between postpran dial insulinemia and the storage and oxidation of fat, protein, and carbohydrate may provide further insight into the relation between fuel metabolism and satiety, and establish whether low-insuline mic diets can facilitate greater body fat loss than isoenergetic high-insWinemic diets.

We thank Efi Farmakalidisfor her assistancein the planningof this study and Natasha Porter for her technical assistance with the experimental work for the carbohydrate-rich food group.

REFERENCES

1. Jenkins DJA, Wolever TMS, Kalmusky J, et al. Low glycemic index

carbohydrate foods in the management of hyperlipidemia. Am J Clin Nutr 1985;42:604â€”17. 2. Jenkins DJA, Wolever TMS, Vuksan V. et al. Nibbling vs gorging:

metabolic advantages of increased meal frequency. N EngI J Med 1989;32l:929â€”34. 3. Bymes SE, Brand Miller JC, Denyer GS. Development of insulin

resistance in rats after low- vs high-amylose diets. Bachelor of Science Honours Thesis, Department of Biochemistry, The University of Sydney, 1993. 4. Slabber M, Barnard HC, Kuyl JM, Dannhauser A, Schall R. Effects of

low-insulin-response,energy-restricted diet on weight loss and plasma insulin concentrations in hyperinsulinemic obese females. Am J Clin Nutr l994;60:48â€”53. 5. Lerer-MetzgerM, Rizkalla SW, Luo J, et al. Effects of long-term

low-glycaemic index starchy food on plasma glucose and lipid con centrations and adipose tissue cellularity in normal and diabetic rats. Br J Nutr 1996;75:723â€”32. 6. Bymes SE, Brand MillerJC, DenyerGS. Amylopectinstarch pro

motes the development of insulin resistance in rats. J Nutr 1995;125:1430â€”7. 7. SalmÃ©ronJ,AscherioA, Rimin EB,etal. Dietaryfiber,glycemicload,

and risk of N!DDM in men. Diabetes Care l997;20:545-50. 8. SalmÃ©ronJ, Manson JE, Meir J, et al. Dietary fiber, glycemic load, and

risk of non-insulin-dependent diabetes mellitus in women. JAMA l997;l2:472â€”7. 9. Modan M, Halkin H, Almog 5, et al. Hyperinsutinemia: a link between

hypertension, obesity and glucose intolerance. J Clin Invest l985;75:809â€”17. 10. Zavaroni I, Bonora E, Pagliara M, et a!. Risk factors for coronary

artely disease in healthy persons with hyperinsulinaemia and normal glucose tolerance. N Engi J Med l989;320:702â€”6.

Downloaded from https://academic.oup.com/ajcn/article-abstract/66/5/1264/4655967 by guest on 26 February 2019

#### Page 13

1276 HOLT ET AL

11. DeFronzo RA, Ferrannim E. Insulin resistance. A multifaceted syn

drome responsible for NIDDM, obesity, hypertension, dyslipidemia, and atherosclerotic cardiovascular disease. Diabetes Care 199l;14:173â€”94. 12. DeFronzo RA, Bonadonna RC, Ferrannini E. Pathogenesis of

NIDDMâ€”abalanced overview. Diabetes Care 1992;15:3l8â€”68. 13. Colditz GA, Manson JE, Stampfer Mi, Rosner B, Willett WC, Speizer

FE. Diet and risk of clinical diabetes in women. Am J Clin Nutr 1992;55:lOl8â€”23. 14. Haffner SM, Valdez BA, Hazuda HP, et al. Prospective analysis of the

insulin-resistance syndrome (syndrome X). Diabetes l992;41:715-22. 15. Lillioja 5, Moft DM, Spraul M, Ct al. Insulin resistance and insulin

secretory dysfunction as precursors of non-insulin-dependent diabetes mellitus. N Engl J Med 1993;329:1988â€”92. 16. StoutR. Hypennsulinaemiaas a riskfactorforcardiovasculardisease.

hit Diabetes Fed Bull 1994;39:l8â€”9. 17. Zavaroni I, Bonini L, Fantuzzi M, et al. Hyperinsulinaemia, obesity,

and syndrome X. J Intern Med l994;235:51â€”6. 18. Wolever TMS, Jenkins DJA, Jenkins AL, Josse RG. The glycemic

index: methodology and clinical implications. Am J Clin Nutr l99l;54:846-54. 19. ServiceFJ,PizzaPA, HallLD,Ctal.Prandialrequirementsin insulin

dependent diabetics: effects of size, time of day, and sequence of meals. J Cliii Endocrinol Metab 1983;57:931-.6. 20. Simpson RW, McDonald J, Wahlqvist ML, Altey L, Outch K. Ma

cronutrients have different metabolic effects in nondiabetics and din betics. Am J Cliii Nutr 1985;42:449â€”53. 21. Nuttall FQ, Mooradian AD, Gannon MC, et al. Effect of protein

ingestion on the glucose and insulin response to a standardized oral glucose load. Diabetes Care l984;7:465â€”70. 22. Krezowski P, Nuttall FQ, Gannon MC, Bartosh NH. The effect of

protein ingestion on the metabolic response to oral glucose in normal individuals. Am J Clin Nutr l986;44:847â€”56. 23. Collier G, McLean A, O'Dea K. Effect of co-ingestion of fat on the

metabolic responses to slowly and rapidly absorbed carbohydrates. Diabetologia 1984;26:50-.4. 24. Gannon MC, Nuttall FQ, Westphal SA, Seaquist ER. The effect of fat

and carbohydrate on plasma glucose, insulin, C-peptide, and triglyc erides in normal male subjects. J Am Coll Nutr l993;12:36-41. 25. Nuttall FQ, Gannon MC. Plasma glucose and insulin response to

macronutrients in nondiabetic and NIDDM subjects. Diabetes Care l991;l4:824â€”38. 26. Morgan L Insulin secretion and the entero-insular axis. In: Rau PR,

ed. Nutrient regulation of insulin secretion. London: Portland Press Ltd. 1992:1â€”22. 27. Teff K. Cephalic phase insulin release in humans: mechanism and

function. In: Fernstrom J, Miller G, eds. Appetite and body weight regulation. Sugar, fat and macronutrient substitutes. London: CRC Press, 1993:37â€”49. 28. Holt SHA, Brand Miller JC, Petocz P. Farmakalidis E. A satiety index

of common foods. Eur J Clin Nutr 1995;49:675â€”90. 29. Wolever TMS, Jenkins DJA. The use of the glycemic index in pre

dicting the blood glucose response to mixed meals. Am J Clin Nutr l986;43:167â€”72. 30. Estrich D, Ravnik A, Schlief G, Fukayama 0, Kinsell L Effects of

co-ingestion of fat and protein upon carbohydrate-induced hypergly cemia. Diabetes 1967;16:232â€”7. 31. Rasmussen 0, Whither E, Arnfred J, Hermansen K. Comparison of

blood glucose and insulin responses in non-insulin dependent diabetic patients. Eur J Clin Nutr l988;42:953â€”61. 32. Peters AL, Davidson MB. Protein and fat effects on glucose responses

and insulin requirements in subjects with insulin-dependent diabetes mellitus. Am J Clin Nutr 1993;58:555-60. 33. Wolever TMS, Katzman-Relle L, Jenkins AL, et al. Glycaemic index

of 102 complex carbohydrate foods in patients with diabetes. Nutr Res l994;l4:651â€”69. 34. The Diabetes Control and Complications Trial Research Group. The

effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus. N EngI J Med 1993;329:977â€”86. 35. Coulston AM, Hollenbeck CB, Swislocki ALM, Reaven GM. Persis

tence of hypertriglyceridaemic effect of low-fat, high-carbohydrate diets in NIDDM patients. Diabetes Care 1989;l2:94â€”lOl. 36. Rivellese A, Giacco R, Genovese 5, et al. Effects of changing amount

of carbohydrate in diet on plasma lipoproteins and apolipoproteins in type II diabetic patients. Diabetes Care 1990;13:446-8. 37. Garg A, Grundy SM, Unger RH. Comparison of effects of high and

low carbohydrate diets on plasma lipoproteins and insulin sensitivity in patients with mild NIDDM. Diabetes l992;4l:1278â€”85. 38. Garg A, Bantle JP, Henry RR, et al. Effects of varying carbohydrate

content of diet in patients with non-insulin-dependent diabetes. JAMA 1994;271:1421â€”8. 39. Campbell LV, Marmot PE, Dyer JA, Borkman M, Storlien LH. The

high-monounsaturated fat diet as a practical alternative for NIDDM. Diabetes Care 1994;17:177â€”82.

Downloaded from https://academic.oup.com/ajcn/article-abstract/66/5/1264/4655967 by guest on 26 February 2019

---

Source: [Fermat's Library](https://fermatslibrary.com/) — https://fermatslibrary.com/s/an-insulin-index-of-foods-the-insulin-demand-generated-by-1000-kj-portions-of-common-foods. More annotated papers: https://fermatslibrary.com/llms.txt
