the food insulin index v2

It’s generally difficult for healthy people to eat too much protein.  However the fact that protein requires some insulin to metabolise is an important consideration for people who need to inject extra to keep their blood glucose levels stable.

A better understanding of the insulin response to various foods would be useful for diabetics calculating their insulin dose or even to help refine food choices to manage insulin load.

Since launching the optimising nutrition blog I have had many interesting discussions and learned a lot about protein and how it affects insulin and blood glucose.

The Most Ketogenic Diet Foods article which reviews the food insulin index data and what we can learn about our food choices has received almost 200,000 view.  Given the level of interest, I thought it would be useful to review this topic in more detail.

the food insulin index… a quick refresher

If you’ve been reading Optimising Nutrition blog you would have come across discussion of the recent food insulin index testing undertaken at the University of Sydney as detailed in Kirstine Bell’s PhD thesis Clinical Application of the Food Insulin Index to Diabetes Mellitus [1] (Sept 2014).

The primary learning from the recently expanded food insulin index data is that the carbohydrate content of a food only partially explains the insulin response.  The protein, fibre and fructose also affect our insulin response to our food.

The cluster of data points on the left-hand side of the figure below shows that:

  1. low carbohydrate, high fat foods trigger a negligible insulin response, while
  2. low carbohydrate high protein foods cause a significant insulin response.


When we assume that fibre is indigestible and protein has about half the insulinogenic effect of carbohydrates we get a much better prediction of insulin response.


The insulin requirement of a particular food is described better by the following formula:

insulin load = total carbohydrates – fibre + 0.56 * protein

digestion time for protein versus carbohydrates

One of the limitations of the food insulin index data is that the insulin area under the curve was measured over only three hours.  This is not a big deal for foods that are high in carbohydrates as they are generally fully digested within three hours.

However protein can take much longer to digest.  In the article The Blood Glucose, Glucagon and Insulin Response to Protein we saw that the insulin response to protein in diabetics can be even greater and over a longer period than for people who do not have diabetes.

If we were to repeat the food insulin index testing over a longer period it is likely that the measured insulin response would be significantly greater and even more-so in people with diabetes.  That is, the insulin response to protein may be greater than the 56% of the insulin response to carbohydrate indicated by the analysis of the food insulin index data if we were to measure the insulin response over a longer period.

Wilder’s ketogenic formula

Dr Russell Wilder of the Mayo Clinic was the first to coin the term ‘ketogenic diet’. [2]  Wilder developed the diet as an alternative to fasting in the treatment of epilepsy in the 1920s.

Image result for dr russell wilder

Wilder also developed the formula shown below to determine whether a diet would be ketogenic.  If the number from this calculation was greater than 1.5 (ideally greater than 2.0) then the diet would be considered to be ketogenic and appropriate for the treatment of epileptics. [3]


This formula is based on the understanding that:

  • 100% of carbohydrate is glucogenic (i.e. converts to glucose),
  • 54% of protein is glucogenic,
  • 46% of protein is ketogenic, and
  • 10% of fat is glucogenic.

I had previously searched for detail of how Wilder had arrived at the 56% / 46% split for protein and only found references suggesting that the 56% glucogenic potential of protein comes from the analysis of nitrogen in the urine of dogs. [4]  However I recently came across this paper which details Wilder’s thinking in more detail.

Wilder’s conclusion that a diet needs to have more than two times the ketogenic precursors compared to glucogenic precursors is still the basis of the formulation of diets used to treat epilepsy.

According to George Cahill, Krebs also found that 57g of glucose may be derived from 100g of protein. [5]   Again, this is similar to the insulin demand for protein observed in the food insulin index tests.

carbohydrate counting

The most straight forward approach is to assume that protein has no impact on insulin or blood sugars.

Dr Richard Berstein and Dr Robert Atkins pioneered the concept of carbohydrate counting for weight loss and diabetes management in the 70s and 80s.  There have been various waves of popularity of low carbohydrate diets with many people finding success.

Carbohydrate counting alone is a reasonable approach that is likely to work for most people, particularly if they are not highly insulin resistant.

However, there are some people that reducing carbohydrates alone doesn’t work for.   The fact that protein also generates insulin suggests that managing protein as well as carbohydrates may be necessary to manage insulin levels.

thermic effect of food

You may have heard of the concept of the thermic effect of food where different foods require different amounts of energy for the digestion process.  For example, a mushroom, which has a very low calorie density and a lot of fibre and protein, may require more energy to digest than is obtained from the digestion of the mushroom.

The maximum and minimum thermic effect (also known as the specific dynamic action) for each macronutrient is shown below. [6]

Microsoft Word Document 3072015 40140 PM.bmp

Compared to carbohydrates and fat, protein only yields between 76% and 84% of the energy per calorie ingested because of losses in digestion.  This is useful to know if you’re trying to minimise calorie intake.

As discussed in the Why We Get Fat V2 article, part of this thermic effect of food is also likely to be due to the fact that there is a significant loss of energy when we convert protein to glucose to be used as energy.  The body doesn’t like to do this other than in an emergency.

Steve Phinney’s “well formulated ketogenic diet”

One of the key observations from Steve Phinney’s well formulated ketogenic diet (WKFD) chart is that we need to strike a balance between carbohydrates and protein in order to maximise the ketogenic potential of our diet.


You can have 30% protein and 5% carbs or 20% carbs and 10% protein and still be within the bounds of a ketogenic diet.  However if you have 30% protein and 20% carbs you will be outside the realms of a ketogenic diet because you will be producing too much glucose.

According to Nuttall and Gannon [7] the body requires between 32 and 46g per day of high quality dietary protein to maintain protein balance.  This equates to around 6 to 7% of calories in a 2000 to 2500 calorie diet being taken ‘off the top’ for growth and maintenance, with everything else potentially available as ‘excess’ protein for gluconeogenesis.  This should not be considered optimal, but simply a minimum reference point for the absolute minimum amount of protein.

Interestingly, the slope of the line along the face of Phinney’s WFKD triangle corresponds with the assumption that 7% of protein goes to muscle growth and repair (protein synthesis) with 75% of the remaining ‘excess’ protein being glucogenic.  This 75% value is in the “ball park” (although a little higher) of our previous estimate of the glucogenic potential of protein based on the analysis of the food insulin index data.

amino acid potential

We also have an understanding of which amino acids are glucogenic, which are ketogenic and which are a bit of both. [8] [9] [10]  The table below shows the various amino acids divided up on the basis of their ketogenic versus glucogenic potential and also which are essential versus non-essential. [11]

Microsoft Word Document 3072015 34915 PM.bmp

Only two amino acids are exclusively ketogenic.  There is a handful that are both glucogenic and ketogenic.  However most of the amino acids are glycogenic, meaning that they will most likely turn into glucose if not required for protein synthesis.

According to David Bender “In  fasting  and  on  a  low  carbohydrate diet  as  much  of  the  amino  acid  carbon  as  possible  will  be  used  for gluconeogenesis, an ATP-expensive, and hence thermogenic process.” 

Hence it appears likely that in a low carbohydrate diet situation excess amino acids that fit under the “both” classification will be turned to glucose rather than ketones because the body needs the extra glucose which it is not getting from ingested carbohydrates.

Conversely, if someone is consuming a high carbohydrate diet the excess amino acids that fit into the “both” category will be converted to ketones rather than glucose because the body is getting more than enough glucose from the diet.

So, to some extent, protein is versatile depending on the body’s need. But at the same time, it is only a small portion of the amino acids that are able to do this. The fate of the majority of the amino acids is pre-destined.

the krebs cycle

The figure below shows the process of catabolism of amino acids. [12]


I am not an organic chemist, but from what I understand this means that:

  • The amino acids Leucine and Lysine cannot be converted back to glucose as they are ketogenic;
  • Isoleucine, Tyrosine, Phenylalanine, Tryptophan, Threonine all enter into the amino acid catabolism cycle and can be used for various functions, such as muscle repair and growth, but can also be converted back into glucose if required (glucogenic) or turned into fatty acids (ketogenic); and
  • The remaining amino acids enter the cycle and can be used for a variety of functions in the body, but cannot be converted into fatty acids.  If they are not required they can be turned into glucose and potentially stored as body fat.

The majority of the amino acids obtained from the digestion of protein have the potential to be turned into glucose through gluconeogenesis.

The reason that we don’t see a sharp rise in blood glucose is partly because amino acids from digestion circulate in the blood until they are required.  Gluconeogenesis is a demand driven process.  Glucose is pulled from amino acids when there is no other source rather than pushed into the bloodstream due to ingestion of excess protein.

By contrast, glucose from carbohydrates will be used to refill glycogen stores (liver and muscle) and then find their way quickly into the bloodstream.  In most people, the amino acid stores in the blood are not saturated and hence there is plenty of capacity to store amino acids until they are required, at least if you have good insulin sensitivity and are not diabetic.

The body does need glucose, and it is fine to get it from carbohydrates or protein via gluconeogenesis.  However many people struggle to produce enough insulin and / or are insulin resistant and hence struggle to keep their blood sugars in normal range.  For these people it makes sense to reduce the insulin load their diet (the portion that requires insulin) to a point that they can maintain normal blood glucose levels.

tallying up the amino acids

I figured I could use this knowledge of the categorisations of the various amino acids to better understand how much of the proteins in the 8000 foods listed in the USDA food database are glucogenic versus ketogenic.

For each food in the USDA database I tallied up the weight of the glucogenic and ketogenic amino acids and the amino acids that fell onto the ‘both’ category and found that:

  • ketogenic amino acids make up only 12% by weight of the total protein across the 8000 foods in the database,
  • glucogenic amino acids comprise 74% of the foods, and
  • amino acids that fit in the “both” comprise 14% of the total weight of amino acids.

This means that somewhere between 78% and 89.5% of protein has the potential to turn into glucose, depending on whether you considered the amino acids in the ‘both’ column to be glucogenic or ketogenic, or somewhere in between.

For someone eating a low carbohydrate diet nearly 90% of ‘excess’ protein could be turned to glucose in the blood stream.

Why is this different to the observation from the food insulin index testing that approximately 56% of protein raises insulin?  Perhaps the following factors come into play:

  1. When we consider the glucogenic potential of the individual amino acids we are considering the maximum potential of protein if it is not first used for protein synthesis.  The amount of protein synthesis will be greater for say an athlete or a body builder, with less protein remaining for gluconeogenesis.
  2. Converting protein to glucose requires energy and hence some of the energy from ingested protein is lost in the process and hence is not converted to glucose.
  3. The insulin index testing is undertaken over only three hours. Protein takes much longer to digest and be metabolised into glucose hence the insulin index testing may underestimate the full glucogenic potential of protein.

which foods have the most ketogenic protein?

So I bet you are wondering which forms of protein have the highest amount of ketogenic protein.  Maybe not?  Well, I was, and I am going to share it with you.

The table below shows the foods from the USDA database that have the most ketogenic protein (assuming the ‘both’ amino acids are split 50/50 glucogenic / ketogenic) in terms of grams of ketogenic amino acids per 100 grams of the food.

Food ketogenic aminos ( per 100g) % ketogenic protein % insulinogenic
Seal, Bearded Alaskan 19.4g 23% 72%
Whale, Beluga 17.6g 25% 64%
Cod 16.3g 26% 68%
Seaweed, spirulina 14.2g 25% 64%
White fish 13.6g 22% 53%
Parmesan cheese 12.3g 32% 28%
Beef, sirloin 10.0g 33% 50%
Beef, ribeye 9.7g 33% 44%
Bacon 9.3g 25% 22%
Egg yolk 9.2g 27% 18%
Lamb 9.0g 25% 39%
Chicken, breast with skin 7.8g 24% 48%
Salmon 7.0g 28% 45%
Egg, whole 3.3g 26% 29%
Milk 0.9g 29% 43%

It is hard to know what to make of this list other than noting that the seal, whale and cod have the highest amounts of ketogenic protein.  Perhaps there is something about cold water animals that cause them to store more ketogenic amino acids?  This seems to align with what we see in the traditional diets of humans who may eat more fat if they are living further away from the equator but eat more carbohydrates from fruits if they live closer to the equator.

Although seal, whale and cod have high amounts of ketogenic amino acids, overall they are still quite insulinogenic.  In view of the high proportion of insulinogenic properties of some meats it is not surprising that people can thrive on a 100% meat zero carb diet because the body can get as much glucose they need from the meat.[13]  At the same time though, I’m not sure that an all meat diet can provide an optimal array of vitamins and minerals unless you are emphasising organ meats.

In view of the fact that a large amount of protein can be converted to glucose through gluconeogenesis, it seems better to focus on foods that have a lower percentage of insulinogenic calories if you are insulin resistant or do not have a fully functioning pancreas.

Rather than worrying about whether you’re eating too much protein, most people will do fine if they limit their processed grains and sugars and eating as much protein as their appetite directs them to.  If you are aiming for a therapeutic ketogenic diet to manage chronic conditions such as cancer, epilepsy or dementia, then you may want to consider moderating your protein intake to drive ketosis.

While there is no such thing as a glycemic index for protein, it also makes sense to avoid processed foods if you are after stable blood glucose levels and lasting satiety.  Unless you are a bodybuilder who is looking for a quick insulin spike it would be prudent to prioritise protein from whole foods.


The table below shows a comparison of a range of glucogenic factors for protein relative to carbohydrate, summarising the discussion above.  Most of the approaches to understanding the insulinogenic portion of protein give an even higher value than suggested by the analysis of the food insulin index data.

Basis % insulinogenic Comment
Carbohydrates only 0% A lower end sensitivity assuming that no protein is converted to glucose (i.e. as per standard carbohydrate counting).
Food insulin index 56% Based on testing of > 100 foods in healthy individuals
Thermic effect of food 77% Average of additional in digestion losses minus 7%.
Wilder’s formula 54% Used in initial ketogenic formula
Krebs  / Janney 57% Based on nitrogen excretion in dogs
Glucogenic potential (min) 78% Based on summing amino acids in USDA foods database, excluding “both” aminos.
Glucogenic potential (max) 89.5% Based on summing amino acids in USDA foods database, including “both” aminos.
Steve Phinney WFKD 75% Assuming that the first 7% of calories goes to growth and repair with 75% of the remaining amino acids being glucogenic.

the most ketogenic foods… updated

I have calculated the insulinogenic potential of the foods shown in this previous article (The Most Ketogenic Diet Foods) using the following approaches:

  • carbohydrates only;
  • food insulin index data (i.e. protein is 56% insulinogenic);
  • thermic effect (i.e. protein is 77% insulinogenic); and
  • maximum glucogenic potential of the amino acids for each food (varies for each food based on data in USDA foods database).

This updated data illustrates the difference in standard carbohydrate counting and the full insulinogenic potential of the food.  While there is a range of values due to the varying amounts and types of protein overall, there is a reasonable alignment between the food insulin index (56%), thermic effect of food (77%) and maximum glucogenic potential values, particularly when we compare it to the carbohydrate only approach for the lowest carbohydrate foods.

least insulinogenic foods

food carb only (0%) FII (56%) thermic (77%) glucogenic (max)
olives 1% 4% 4% 4%
cream 3% 4% 6% 4%
pecans 2% 5% 8% 6%
Macadamia nuts 3% 5% 7% 6%
duck 0% 7% 4% 9%
pork sausage 2% 10% 19% 9%
sesame seeds 7% 7% 10% 11%
sausage 0% 9% 12% 14%
frankfurter 2% 11% 14% 14%
pepperoni 0% 10% 14% 15%
bacon 1% 16% 21% 21%
mackerel 0% 20% 28% 28%


egg  carb only (0%) FII 56%) thermic (77%) glucogenic (max)
egg yolk 16% 15% 20% 19%
whole egg 17% 21% 23% 25%
egg white 6% 53% 71% 72%

Dairy products


cheese carbs only (0%) FII (56%) thermic (77%) glucogenic (max)
cream cheese 5% 9% 10% 9%
brie 1% 14% 20% 18%
limburger 1% 14% 19% 18%
camembert 1% 15% 21% 19%
Monterey 1% 15% 20% 19%
cheddar 1% 15% 20% 19%
gruyere 0% 17% 23% 20%
Colby 3% 16% 21% 20%
blue 3% 16% 21% 20%
edam 2% 17% 23% 21%
gouda 2% 18% 24% 22%
feta 6% 18% 23% 22%
ricotta, whole milk 7% 21% 27% 24%
mozzarella 3% 20% 26% 26%
cream cheese, low fat 16% 25% 28% 27%
parmesan 3% 21% 27% 28%
mozzarella, skim milk 4% 26% 34% 31%
Swiss 6% 22% 27% 34%
ricotta, part skim milk 15% 33% 40% 37%
cream cheese, fat free 29% 62% 75% 72%
Swiss, low fat 8% 45% 48% 73%
cottage cheese, low fat 17% 55% 69% 86%
mozzarella, non-fat 10% 60% 79% 95%


milk carb only (0%) FII (56%) thermic (77%) % insulinogenic (max)
Full cream milk, 3.7% fat 29% 41% 41% 43%
Human milk 40% 43% 44% 43%
Skim milk, 1% fat 47% 65% 72% 69%
Chocolate milk, low fat 63% 72% 76% 70%


yogurt carb only (0%) FII (56%) thermic (77%) % insulinogenic (max)
plain, whole milk 30% 42% 48% 46%
Plain, low fat 44% 63% 70% 68%
fruit, low fat 71% 81% 85% 83%
plain, skim milk 55% 78% 87% 85%
fruit, non-fat 70% 90% 97% 96%


fruit carb only (0%) FII (56%) thermic (77%) % insulinogenic (max)
olives 1% 3% 4% 4%
avocados 4% 8% 9% 7%
raspberries 42% 42% 51% 45%
blackberries 40% 42% 53% 47%
strawberries 70% 75% 76% 69%
oranges 77% 81% 83% 76%
apples 88% 89% 89% 81%
bananas 91% 91% 95% 86%


vegetable carb only (0%) FII (56%) thermic (77%) % insulinogenic (max)
endive 6% 22% 29% 24%
dock 5% 27% 33% 27%
mustard greens 7% 61% 43% 34%
asparagus 36% 60% 69% 34%
artichoke 22% 35% 39% 38%
sauerkraut 30% 41% 45% 40%
broccoli 3% 35% 47% 42%
lettuce 28% 44% 50% 42%
coriander 15% 36% 44% 43%
chrysanthemum leaves 0% 32% 43% 44%
alfalfa 3% 42% 57% 47%
parsley 34% 52% 59% 48%
cauliflower 32% 50% 56% 48%
spinach 24% 53% 63% 50%
bamboo shoots 19% 50% 62% 51%
mushroom 31% 56% 66% 55%
turnip 17% 30% 34% 62%
onions 78% 85% 88% 82%

Nuts, seeds and legumes

nuts, seeds legumes carbs only (0%) FII (56%) thermic (77%) % insulinogenic (max)
pecans  2% 5%  10% 5%
Macadamia  3% 5%  6% 6%
coconut meat  7% 6%  10% 7%
coconut cream  6% 7% 9% 8%
coconut milk  6% 7% 9% 8%
Brazil nuts  3% 7% 10% 9%
flax seed  1% 8% 12% 11%
walnuts  4% 9%  11% 11%
pine nuts  5% 9%  11% 11%
sesame butter (tahini)  6% 11% 15% 14%
sesame seeds  0% 12% 10% 15%
chia seeds  6% 13% 17% 16%
peanuts  4% 13%  19% 18%
sunflower seeds  9% 14% 19% 18%
pumpkin seeds  6% 14% 22% 19%
pistachio nuts  12% 19%  23% 22%
cashew butter  21% 22% 29% 25%
almonds  7% 13% 18% 17&


fish carbs only (0%) FII (56%) thermic (77%) % insulinogenic (max)
Tuna 0% 32% 44% 44%
Mackerel 0% 33% 46% 25%
Herring 0% 19% 26% 25%
Salmon 0% 24% 33% 34%
Sardine 0% 26% 36% 36%
Anchovy 0% 31% 42% 42%
Swordfish 0% 31% 42% 42%
Trout 0% 31% 44% 43%
Carp 0% 32% 43% 43%
Yellowtail 0% 36% 49% 49%
Bass 0% 37% 51% 51%
Mullet 0% 37% 51% 51%
Squid 18% 41% 49% 51%
Abalone 23% 47% 55% 57%
Monkfish 0% 44% 59% 60%
Halibut 0% 47% 24% 61%
Mussel 17% 49% 60% 62%
Oyster 21% 46% 56% 63%
Crab 0% 48% 66% 65%
Shrimp 5% 48% 64% 65%
Hadock 0% 51% 68% 66%
Perch 0% 49% 65% 67%
Clam 14% 56% 67% 71%
Scallop 19% 59% 76% 80%


meat carbs only (0%) FII (56%) thermic (77%) % insulinogenic (max)
Bologna 6% 12% 17% 14%
Frankfurter 2% 11% 14% 14%
Duck 0% 14% 17% 17%
Chorizo 2% 15% 18% 17%
Beef, ribeye 0% 15% 26% 21%
Bacon 1% 15% 21% 21%
Pork, ham 6% 17% 38% 22%
Pork, blade, hocks & shoulder 31% 23% 42% 31%
Turkey 0% 23% 29% 32%
Lamb mince 0% 24% 27% 34%
Chicken 0% 24% 34% 34%

post last updated May 2017






[5] – Cahill references a 1964 paper by Krebs in this paper but I can’t find the original paper.










19 thoughts on “the food insulin index v2”

  1. Will there be changes made to the calculator now in light of this new info? Is there an insulinogenic calculator app (or one in the works), so smartphone users can check on foods right there in the grocery store aisle?

    Hubby will miss his ribeye, as it’s now outside his tolerance zone.


    1. I’m not sure we really need to update the previous calculator as I think the 56% is a pretty good average for most people. I suppose what this analysis has taught me is that 56% is not nuts as the upper limit could be even higher.

      Some type 1 diabetics will use even higher ratios to calculate their insulin dose for protein based on their own testing.

      Poor hubby. No ribeye. 😦 Perhaps you should try some and see how his blood sugar reacts in the hours afterwards. Or maybe fattier cuts of meat?

      My intention is not to encourage people to not eat protein, but rather to work within their own personal tolerance limits for carbs and protein. The insulin load formula will encourage you to reduce carbs. increase fibre and then moderate protein if your are insulin resistant.

      Protein is important. More discussion on that to come.


      1. While we were eating dinner last night (one last celebratory ribeye), I had a thought: wait a minute–we eat grass-fed/pastured meats. Would this have an effect on the new insulin index info, since what you have is based in NDB data (basically CAFO meats)? Can Hubby have his ribeye back? 🙂


      1. I’m with you Walter. In the end it’s about the insulin load (i.e. the cumulative impact of all the food you eat) rather than the insulinogenic properties of just one meal. Steak is certainly insulogenic but it’s not off the chart. Most people would be able to handle a reasonable size steak without causing too much of an extreme blood glucose excursion. Seems that people with more advanced diabetes may struggle with their glucagon response and will see an early rise in blood sugars due to a large protein meal. The only way to tell is to test the blood glucose to an actual meal and refine. ‘Eat to your metre.’


    1. Thank you from those who can now use their smartphones right there in the grocery store. I personally don’t own a smartphone, but now I might have to go out and buy one…or maybe I can just have Hubby enter your mathematical formula into his engineering calculator, and see if that works.

      Liked by 1 person

  2. This is the stuff I practically raped from your site–it’s now all over my fridge held on with magnets. Hubby’s calculator is non-programmable, but he very kindly set me up an Excel sheet with the new formula built into it. Not exactly mobile, but more accurate for his needs. Meanwhile, we have some small electronic device shopping to do.

    Bio-hacking Hubby has been a very long N=1 experiment, and I grab help wherever I can find it (to avoid ending up spending our retirement money on insulin he can get by without–for now).

    Liked by 1 person

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s