BEYOND GLYCEMIC INDEX:

NEW FOOD INSULIN INDEX 
Laura Sampson Kent*
Mary Franz*
Jennie Brand Miller**
Walter Willett*

*Department of Nutrition, Harvard School of Public Health
**University of Sydney

Acknowledgement: This research was funded by grant CA55075, National Cancer

Institute, National Institutes of Health.
 RELEVANCE OF INSULIN SECRETION

• Prevention and management of weight gain

• Hyperlipidemia 

• Non‐insulin dependent diabetes mellitus (NIDDM)

• Long‐term exposure to high insulin concentrations       
may be associated with increased risk of cancer 
y
 WHAT IS FOOD INSULIN INDEX (FII)?
WHAT IS FOOD INSULIN INDEX (FII)?
Directly quantifies the postprandial insulin 
Directly quantifies the postprandial insulin
response to an isoenergetic portion of a 
test food in comparison to a reference
test food in comparison to a reference 
food 
Allows testing of foods with no or low
Allows testing of foods with no or low 
carbohydrate content
FII has been shown to predict the relative 
h b h d h l
insulin demand evoked by mixed meals 
 FOOD INSULIN INDEX (FII)
FOOD INSULIN INDEX (FII)
• Measures
Measures blood insulin response to a food 
blood insulin response to a food
compared to a reference food glucose (FII=100)

• Measures incremental insulin area under the 
curve (AUC) over two hours in response to
curve (AUC) over two hours in response to 
consumption of a 1000 kJ portion of the test 
food divided by the AUC after ingestion of a
food divided by the AUC after ingestion of a 
1000 kJ portion of the reference food
How is FII Different from Glycemic Index?
How is FII Different from Glycemic Index?

FOOD INSULIN INDEX
FOOD INSULIN INDEX GLYCEMIC INDEX
GLYCEMIC INDEX
• Measures postprandial  • Measures effect of 
increase in insulin secretion  carbohydrate‐containing 
of a whole food  foods on postprandial 
increase in blood‐glucose 
• Dependent on 
d levels
levels 
carbohydrate, quantity and  • GI is not always 
quality of protein and fat 
q y p proportional to the insulin 
and their interactions response
 METHODS
• APPROXIMATELY
APPROXIMATELY 100 FOODS AND 16 CEREALS 
100 FOODS AND 16 CEREALS
WERE ANALYZED USING GLUCOSE AS THE 
REFERENCE FOOD
• 407 FOODS, 147 MARGARINES, 58 OILS, 111 
CEREALS WERE IMPUTED USING VARIOUS 
ALGORITHMS
• 166 FOODS AND 89 CEREALS WERE RECIPE‐
DERIVED
• 44 FOODS WERE ASSUMED TO BE 0
 FOODS SENT TO UNIVERSITY OF SYDNEY
FOODS SENT TO UNIVERSITY OF SYDNEY

MUFFIN MIXES • SunMaid honey raisin 
bran muffin mix
• Duncan Hines blueberry 
D Hi bl b
muffin mix
READY‐MADE 
READY MADE • Krusteaz fat‐free 
Kr stea fat free
SWEET ROLL blueberry muffin mix
• Entenmann cinnamon 
Entenmann cinnamon
buns
Cake Mix
Cake Mix • Betty Crocker golden cake w/ 
Betty Crocker choc frosting
Pancake Mix • Aunt Jemima pancake mix

• Chips Ahoy regular chocolate chip 
Cookies • Chips Ahoy reduced fat chocolate 
Chips Ahoy reduced fat chocolate
chip 
• Archway fat‐free oatmeal raisin 

• Cape Cod reduced fat potato 
Chips
p chips
Candy Bars • Snickers candy bars
• Hershey candy bars
y y
 U.S. Analyzed Cereals
U.S. Analyzed Cereals
• Cheerios
CEREALS • Shredded Wheat
• Corn Flakes
• Quaker 100% Natural
• Cracklin Oat Bran
• L k Ch
Lucky Charms
• Frosted Flakes
• Great Grains
Great Grains
• Honey Bunches of Oats
• Wheaties
 Matching U.S. Foods Purchased in Sydney
Dairy Meat, fish, mixed dishes
• Cream cheese • Tacos
• Low fat cheddar cheese • Lasagna
• Low fat processed cheese • Roast chicken
• Low fat cottage cheese • Shrimp
• Low fat ice cream • Tofu
• Sherbet • Tuna in oil
Tuna in oil
• 1% milk • Bacon
• Butter
Miscellaneous items
F it
Fruits or fruit drinks
f it d i k • Olive oil
• Peaches canned in juice • Navy beans
• Peaches canned in syrup • Raspberry jam
• Orange juice
g j • Pretzels
• Prunes • Walnuts
• Punch • Corn tortillas
Cereals • Jatz crackers (similar to Ritz)
• Kashi 7 grain puffs
K hi i ff
• Rice Bubbles
 Recipe‐derived
Recipe derived FII Values 
FII Values

• Used analyzed foods as ingredients
• Used ingredients that were imputed from 
analyzed foods.
y
Example: apple butter                  FII = 52
Derived from recipe using apple, a.j., and 
df l d
sugar as ingredients (all analyzed foods)
 IMPUTED OR CALCULATED FII VALUES
IMPUTED OR CALCULATED FII VALUES
FOODS CALCULATION

Cereal grains, baked goods,  Similar analyzed foods and a 
sweets and snack foods,  carb/1000 kj adjustment
beverages fruits and dairy
beverages, fruits, and dairy
carb/1000 kj imputed food x FII
carb/1000 kj analyzed food

Breakfast cereals Used analyzed cereals adjusting for 
carbohydrate/amount

Meats, fish, and poultry Directly from similar analyzed foods
 IMPUTED FII VEGETABLE VALUES
IMPUTED FII VEGETABLE VALUES
VEGETABLES CALCULATIONS
• NON STARCHY Recipe from avocado using an 
h2o ratio

• STARCHY Recipe from potatoes or 
carrots using fiber ratios
carrots using fiber ratios

Recipe from beans using an 
• PEAS and LIMA BEANS fiber ratio
 FII PUBLICATIONS
FII PUBLICATIONS
Bao J, de Jong
, g V, Atkinson F, Petocz
, , P, Brand‐Miller 
,
JC. Food insulin index: physiologic basis for 
predicting insulin demand evoked by composite 
meals Am J Clin Nutr 2009;90:986‐992.
meals. Am J Clin 2009;90:986 992

Holt SH, Miller JC, & Petocz
Holt SH Miller JC & Petocz P. An insulin index of foods: 
P An insulin index of foods:
the insulin demand generated by 1000‐kJ portions of 
common foods. Am J Clin Nutr, 66:1264‐67; 1997.
 RESULTS of the mixed meal study: 
M7
A Group 1 M1 B Group 2
M8
M2
M9
500 M3 500
M10
M4 M11
400 400
Insulin (pmol/L)

M5 M12
M6 M13
300 300
White Bread White Bread

200 200

100 100

0 0
0 15 30 45 60 90 120 0 15 30 45 60 90 120
Tim e (m in)
Tim e (m in)

Bao J et al. Am J Clin Nutr 2009;90:986-992.

 RESULTS: Correlations
A 140
r = 0.78, P = 0.0016
(Rela tive white bread =10 0) 120
Obse rved insulin responsse

100

80

60

40

20

0
0 20 40 60 80 100 120
Insulin demand predicted by Food Insulin Index
 RESULTS: Correlations
B
140
r = 0.53, P = 0.064
n responsee
120
ve white brread =100))

100
ed insulin

80

60
Observe
(Relativ

40

20

0
0 20 40 60 80 100
Available
a ab e ca
carbohydrate
bo yd ate co
content
te t (g)
 Highest versus Lowest FII Values
Highest versus Lowest FII Values
• Highest
Highest analyzed FII
analyzed FII • Lowest
Lowest analyzed FII
analyzed FII
Jelly Beans (120) Olive Oil (3)
Pancakes (112)
Pancakes  (112) Avocado (5)
Avocado (5)
Honeydew Melon (95) Walnuts (6)
Potatoes (90)
Potatoes (90) Tuna (16)
Tuna (16)
Breakfast Cereals  (70‐113) Chicken (20)
MEAN ANALYZED VS MEAN IMPUTED FII DATA

100
90
80
70
60
50
40
30
20
10
0
ES

Y
N

TS

S
IR

UI

AL
OD
EI
M

EE
DA

FR
OT

RE
GU

 FO

SW

CE
PR
 LE

T 
CH
 &

AS
ANALYZED
AR
TS

KF
ST
NU

EA
IMPUTED
BR
 LIMITATIONS
ƒ Although
Although imputed FII data are similar to 
imputed FII data are similar to
analyzed data, algorithms used for imputing 
FII have not been validated.

ƒ FII
FII variability within some food groups 
variability within some food groups
complicates the assignment of imputed 
values.  For example, analyzed FII for dairy 
foods range from 18 for cream cheese to 86 
for fruited yogurt.
 CONCLUSION
ƒ Most vegetables produce negligible insulin response, 
g p g g p ,
as measured by FII.

ƒ Even fats, and protein sources produce largely 
variable insulin responses, as measured by FII.

ƒ Imputed FII values track closely to analyzed values.

ƒ Refined cereals, sweets, and potatoes produce the 
greatest insulin response, as measured by FII.
 DERIVED INSULIN LOAD
DERIVED INSULIN LOAD

Insulin Load, Glucose = (calories*FII)/100      
• the individual average insulinogenic load 
th i di id l i li i l d
(IL) during the past year was estimated 
f
from FFQs by multiplying the FII of each 
FFQ b lti l i th FII f h
food by its energy content and the 
consumption frequency and summing 
ti f d i
over all reported food items. 
 DIETARY INSULIN INDEX, DII
DIETARY INSULIN INDEX DII

The average dietary DII was 
calculated by dividing the 
average Insulin Load, IL, by the 
I li L d IL b th
total daily energy intake
total daily energy intake.
 RESULTS FROM COHORT STUDIES
RESULTS FROM COHORT STUDIES

MEN WOMEN
• Median IL was 840
Median IL was 840 • Median IL was 677
Median IL was 677
• Median DII was 41.7  • Median DII was 42.7 
TOP 8 FOODS CONTRIBUTING TO IL
TOP 8 FOODS CONTRIBUTING TO IL
MEN WOMEN
Cold Cereal (6.1%) Mashed Potatoes (6.2%)
Potatoes (5 7%)
Potatoes (5.7%)  Skimmed Milk (5 7%)
Skimmed Milk (5.7%)
Dark Bread (4.4%) Cold Cereal (5.3%)
Skimmed Milk (3 2%)
Skimmed Milk (3.2%)  D k B d (4 6%)
Dark Bread  (4.6%)
Bananas (3.2%)  Beef (3.6%)
English Muffin (3 1%)
English Muffin (3.1%)  Y
Yogurt (3.3%)
(3 3%)
White Bread (2.9%) 
White Bread (3.1%) 
Orange Juice (2.8%)
Orange Juice (2.8%) 
English Muffin (2.7%) 
TOP 4 FFQ ITEMS explaining the inter‐
individual variation of insulin load
d d l f l l d
MEN WOMEN
49% 45%
LIQUOR  LIQUOR
COLD CEREAL  WINE
BEER YOGURT
YOGURT 
WINE COLD CEREAL
 COHORT STUDY RESULTS
COHORT STUDY RESULTS
In multivariate adjusted regression models, dietary II and IL were not 
significantly associated with plasma C‐peptide.
i ifi tl i t d ith l C tid

Participants in the highest quintile of both II and IL had 26% higher 
triacylglycerol concentrations than participants in the lowest
triacylglycerol concentrations than participants in the lowest 
quintile (ptrend <0.0001). 

The positive association between II and IL and plasma triacylglycerol 
The positive association between II and IL and plasma triacylglycerol
was strongest in obese (BMI ≥30 kg/m2) participants (difference 
highest versus lowest quintile in II  = 72%, ptrend=0.01). 

An inverse association between DII and HDL cholesterol was observed 
among obese participants (difference = ‐22%, ptrend=0.01 for II). 

DII and IL were not associated with LDL cholesterol, CRP or IL‐6. 
 CONCLUSION
DII and IL were not associated with markers
of glycemic control
control, at least in the fasting
state, but may be physiologically relevant to
plasma lipids
lipids, especially in the obese
obese.