Jason Fung in South Africa

Jimmy Moore and Andreas Einfeldt (Diet Doctor) have both called Dr Jason Fung as the star of the recent low carb conference in Cape Town.  He’s been a real inspiration and encouragement to me too.

Check out there summary of his talk in Cape Town here.

I also highly recommend checking out his series of YouTube videos and his blog Intensive Dietary Management.

Excess insulin is not your friend!  Even if you’re using insulin and medications to manage your blood sugar the long term health prognosis is not great if you’re still pounding the processing carbohydrates that keep the insulin levels high.

is sugar really toxic?

You may have noticed a lot of people quitting sugar [1] or saying sugar is toxic. [2]

But what does the food insulin index data have to say about sugar?  Is it any different to other forms of carbohydrate?

The chart below of sugar content versus insulin demand indicates that sugar plays some role in insulin, however the correlation is weak, at least compared to carbohydrate.  If we want to manage our insulin load we’re probably best to consider our total carbohydrates rather than isolating sugar alone as the only problem.

Microsoft Word Document 14052015 105138 AM.bmp

I ran a correlation analysis on the food insulin index data to see if sugar had a unique effect on insulin compared to non-sugar carbohydrate.

The data suggests that sugar does not generate more insulin than other forms of carbohydrates.  If anything sugar requires slightly less insulin on a gram for gram basis compared to carbohydrates.

This could be because sugar is metabolised quickly and the body pushes out a short burst of insulin to clear the sugar from the blood rather than a long persistent effort which might be the case for a lower glycemic index carbohydrate.


This is not to say that sugar is good for you.  There are obvious issues with consuming significant amounts of sugar including:

  1. Sugar has no fibre and has a very high calorie density so it is not filling and you can end up eating lots of it without feeling full.
  2. Refined sugar has a very low nutrient density, and your body is left searching for nutrition in more food and hence will be prone to over consume calories.
  3. Sugar will cause your blood sugar to rise quickly, your body will produce a burst of insulin which will cause your blood sugar to subsequently crash after the insulin surge and leave you feeling hungry again, craving more sugar to make you feel ‘right’ again.

By contrast, the carbohydrates in non-starchy veggies (e.g. spinach, kale, avocado, asparagus) come packaged with fibre, digest slowly and will leave you feeling full, raise your blood sugar gently and are very hard to overeat. [3].

If you have some form of metabolic dis-regulation (e.g. diabetes, obesity etc) then you need to be thinking about everything that raises your insulin, not just sugar.

[next article… glycemic load versus insulin load…  which one is best?]

[this post is part of the insulin index series]

[Like what you’re reading?  Skip to the full story here.]

[1] https://iquitsugar.com/

[2] https://www.youtube.com/watch?v=dBnniua6-oM

[3] I’m not going to deliver into the fructose / glucose issue.  If you want to go there check out https://www.youtube.com/watch?v=dBnniua6-oM or the condensed Shaun Croxton version at https://www.youtube.com/watch?v=tdMjKEncojQ

baked creamed spinach

This baked creamed spinach recipe is from Carrie Brown’s Maramalade and Mileposts blog.

I sort of feel like I know Carrie after listening her host the SANE Shown with Johnathan Bailor.

Even though I’ve analysed nearly 200 meals, anything with spinach in it seems to keep coming to the top.

This one involves spinach, coconut oil, coconut milk, eggs and cream.

This recipe is unique in that it also includes konjac flour or glucomannan flour which has been shown to be an effective prebiotic fibre which fills you up by expanding significantly in the stomach while involving very few calories.

If you wanted to reduce the carb content further you could reduce the onions and coconut cream, though this would likely affect the palatability and enjoyment of the meal.

We haven’t tried this one at our place yet, but I think it looks like a nice change from the typical low carb fare.

Chrome Legacy Window 19042015 42656 PM.bmp

net carbs

insulin load carb insulin fat protein


17g 22g 76% 70% 8%


is the insulin response to protein dose dependent?

The observation that protein requires insulin initially appears to conflict with a number of studies and anecdotal evidence that suggest protein does not have a significant effect on blood sugar. [1] [2] [3]

I think there are two issues to consider.

Firstly, a healthy non-diabetic will be able to deal with the blood glucose rise caused by gluconeogenesis due to protein, hence the overall blood sugar rise may not be significant.  The extension of this argument is that almost any level of protein intake is a good thing.  I haven’t however seen any real discussion around on the insulin effects of eating a diet high in protein diet.

However type 1 diabetics certainly do see a rise in their blood sugar levels that they need to cover with insulin.  Without insulin to blunt the glucogenic effect of the protein, the blood sugar rise from the fast acting protein is not dissimilar to what you would see  from carbohydrates.

The picture below shows how a well controlled type 1 will give small additional boluses as they see they blood sugar rising due to protein.  This is sometimes called “sugar surfing”.  While this approach provides good blood sugar control wouldn’t it be even better if a type 1 diabetic could better predict the insulin requirements before the meal to proactively predict the blood sugar rise rather than just reacting to the blood sugar roller coaster?


If you’re metabolically healthy  the blood sugar rise and insulin secretion due to gluconeogenesis after a large protein meal may not be cause for concern.  However if you are not metabolically healthy and / or are aiming for nutritional ketosis moderating protein to manage excessive gluconeogenesis may be something you want to do.

Secondly, the degree of gluconeogenesis appears to be dose dependent.  If you exercise intensely and the amount of protein you eat is moderate then your body will likely shuttle protein to your muscles for growth and repair as the highest priority.

If you are not active, and you eat a large amount of protein then excess protein will be converted to glucose in your blood stream, raise insulin levels and be sorted as fat.

So while the effect of protein on your blood sugars and insulin is dependent on a number of factors, allowing for about half the insulinogenic effect of carbohydrates from protein appears to be a reasonable starting point..

[next article…  is sugar really toxic?]

[this post is part of the insulin index series]

[Like what you’re reading?  Skip to the full story here.]

[1] http://caloriesproper.com/dietary-protein-does-not-negatively-impact-blood-glucose-control/

[2] http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4342171/pdf/IJE2015-216918.pdf

[3] http://www.ketotic.org/2013/01/protein-gluconeogenesis-and-blood-sugar.html

how long does it take to digest protein?

One of the limitations of the food insulin index data is that the tests were undertaken over a period of three hours, while protein takes a lot longer to fully digest.

As shown in the chart below (from Andreas Einfeldt / Diet Doctor) simple carbohydrates cause blood sugar to rise and fall quickly, however slower digesting protein causes a rise in blood sugar (in a healthy non-diabetic) between four and six hours after a meal (green line).


One of the challenges for type 1 diabetics is that, even if they limit their carbohydrates, their blood sugar will often spike a number of hours after a high protein meal.  And sometimes faster digesting proteins such as protein powder raise the blood sugar much faster than a slow digesting steak.

The image below shows the continuous glucose monitor plot of a type 1 diabetic after a protein shake (46.8g protein and only 5.6g of carbs).  Without insulin to blunt the glocogenic effect of the protein, the blood sugar rise from the fast acting protein is not dissimilar to what you would see from carbohydrates.


On the positive side, protein does not spike blood sugar as much as carbohydrate and is therefore easier to manage.  Though for type 1 diabetics it is important to be conscious of the amount of protein in the diet and manage it accordingly, particularly if they have their carbohydrate intake dialed in and want to achieve optimal blood sugar control.

Similarly, it’s important for type 2 diabetics and people trying to lose weight via a controlled carbohydrate carbohydrate diet to keep in mind that excess protein, although it might not have a significant effect on their blood sugars, will also raise their insulin levels and work against weight loss or nutritional ketosis.

[next article…  is the insulin reaction to protein dose dependent?]

[this post is part of the insulin index series]

[Like what you’re reading?  Skip to the full story here.]

putting it all together… protein and net carbs

So far we’ve learned that carbohydrate alone isn’t a fantastic predictor of insulin requirement.


The observation that protein requires about half as much insulin as carbohydrate improves our estimation of insulin demand.

Then understanding that fibre neutralises the insulin effect of carbohydrates also helps us predict the amount the amount of insulin required by a particular food.

Microsoft Word Document 25032015 40944 AM.bmp

Using net carbohydrates with an allowance for about half the protein gives us a better way to estimate insulin requirement of food compared to using carbohydrates alone.


In order to help us compare various food options we can calculate the proportion of insulinogenic calories of our foods using this formula:


And if we want to keep track of the insulinogenic load of our diet too keep our blood sugars under control or to maintain or achieve nutritional ketosis we can use this formula:


This deeper understanding of the impact of the influence of carbohydrates, protein and fibre may also be useful when it comes to choosing foods with a lower insulin load or even more accurately calculating insulin dosages for diabetics.

[next article…  how long does it take to digest protein?]

[this post is part of the insulin index series]

[Like what you’re reading?  Skip to the full story here.]

how do you like your veggies… cooked or raw?

So what effect does cooking have on veggies?  Is raw better than cooked when it comes to fibre and insulin demand?

To help us understand the effect of cooking the table below shows a comparison of the fibre in a selection of one hundred grams of cooked and raw vegetables.





fibre % insulin carbs fibre

% insulin

spinach 4 2 40% 4 2 40%
broccoli 7 3 54% 7 3 49%
eggplant 6 3 40% 8 3 63%
artichoke 11 5 49% 12 9 29%
mushroom 5 1 55% 5 2 44%
carrots 10 3 66% 8 3 60%

Spinach and broccoli when cooked don’t seem to lose a lot of their fibre.

You could eat 600g of spinach or 300g of broccoli and still have a Bernstein-compliant lunch of dinner (i.e. no more than 12g of carbs).

Eggplant seems to lose some fibre relative to carbohydrates and ends up with an increased percentage of insulinogenic calories.

If you boil something to a mush then it’s probably not going to have the same quality of fibre as if you were to eat it in its raw unprocessed form.  It’s also going to be easier to eat a lot more cooked veggies than lightly steamed or raw veggies.

Lightly steamed is probably your best bet to retain the nutrients and fibre in your veggies.  If you want to check out how your favourite veggies fare before and after cooking you can find out at nutritiondata.self.com.

In view of the growing body of research showing that fibre is good for gut health which is in turn good for diabetes, insulin sensitivity and a whole host of other issues I think it’s hard to build a strong case for avoiding vegetables altogether just to minimise carbohydrates.

[next article…  putting it all together… protein and net carbs]

[this post is part of the insulin index series]

[Like what you’re reading?  Skip to the full story here.]

how much insulin is required to cover protein?

Given that protein appears to contribute to insulin demand I ran a number of scenarios with the food insulin index data to see if insulin requirement is better predicted by carbohydrate in a food plus some proportion of the protein.

Microsoft Word Document 14052015 104549 AM.bmp

The analysis indicates that insulin demand is related to carbohydrate about 60% of the protein.

There’s not a lot of information on the split between glucogenic amino acids and ketogenic amino acids out there, however it seems that only leucine and lysine are exclusively ketogenic and cannot be converted into sugar, while isolucine, threonine, phenylaline, tyrosine and tryptophan are both ketogenic and glucgoenic.  The remaining thirteen of the twenty one amino acids are exclusively glucogenic, meaning that they can be converted to sugar.

The proportion of protein that can turn to glucose relates to the amount of excess protein to the body’s needs, so it will be affected by a number of factors including a person’s activity levels, how much protein and carbohydrates they eat.

The correlation of food insulin index with carbohydrate about half the protein is better than carbohydrate alone (R2 = 0.435 compared to R2 = 0.461) and we no longer have the issue of high protein foods sitting on the vertical axis as shown below.

Microsoft Word Document 14052015 80253 AM.bmp

Accounting for protein in addition to carbohydrate seems to better predict insulin demand.

So in summary, while protein doesn’t spike blood sugar as much as carbohydrates, protein does still require a significant amount of insulin.  People not achieving the desired results from carbohydrate restriction alone may benefit from moderating their protein intake.

[next article…  fibre… net carbs or total carbs?]

[this post is part of the insulin index series]

[Like what you’re reading?  Skip to the full story here.]

superfoods for diabetes & nutritional ketosis

More than carbohydrate counting or the glycemic index, the food insulin index data suggests that our blood glucose and insulin response to food is better predicted by net carbohydrates plus about half the protein we eat.

The chart below show the relationship between carbohydrates  and our insulin response. There is some relationship between carbohydrate and insulin, but it is not that strong, particularly when it comes to high protein foods (e.g. white fish, steak or cheese) or high fibre foods (e.g. All Bran).

food insulin index table - fructose analysis v2 21122015 44912 PM.bmp

Accounting for fibre and protein enables us to more accurately predict the amount of insulin that will be required to metabolise a particular food.  This knowledge can be  useful for someone with diabetes and / or a person who is insulin resistant to help them calculate their insulin dosage or to chose foods that will require less insulin.


If your blood glucose levels are high you are likely insulin resistant (e.g.  type 2 diabetes) or not able to produce enough insulin (e.g. type 1 diabetes) it makes sense to reduce the insulin load of your food so your pancreas can keep up.

This list of foods has been optimised to reduce the insulin load while also maximising nutrient density.  These low insulin load, high nutrient density foods will lead to improved blood sugar control and normalised insulin levels.  Reduced insulin levels will allow body fat to be released and be used for energy to improve body composition and insulin resistance.

As shown in the chart below this selection of foods is also nutrient dense and provides a substantially greater amount of nutrients compared to the average of all foods available.

2017-02-27 (2).png

From a macronutrient perspective these foods have a similar protein content to the rest of the foods in the USDA database, more fibre but much less digestible non-fibre carbohydrate.  And the carbohydrates that are there come from nutrient dense veggies that are hard to overconsume compared to the processed nutrient poor carbs that are typically causing the issues for people.

2017-02-27 (3).png

Included in the table are the nutrient density score, percentage of insulinogenic calories, insulin load, energy density and the multicriteria analysis score (MCA) that combines all these factors.

vegetables and fruit


food ND % insulinogenic insulin load (g/100g) calories/100g MCA
endive 17 23% 1 17 1.9
chicory greens 15 23% 2 23 1.8
alfalfa 12 19% 1 23 1.7
escarole 14 24% 1 19 1.7
coriander 14 30% 2 23 1.6
spinach 19 49% 4 23 1.3
curry powder 5 13% 14 325 1.3
beet greens 12 35% 2 22 1.3
basil 18 47% 3 23 1.3
zucchini 14 40% 2 17 1.3
asparagus 17 50% 3 22 1.2
paprika 8 27% 26 282 1.2
mustard greens 8 36% 3 27 1.1
parsley 14 48% 5 36 1.1
turnip greens 12 44% 4 29 1.1
banana pepper 7 36% 3 27 1.0
collards 8 37% 4 33 1.0
arugula 12 45% 3 25 1.0
lettuce 14 50% 2 15 1.0
chard 14 51% 3 19 1.0
eggplant 5 35% 3 25 1.0
pickles 8 39% 1 12 1.0
cucumber 8 39% 1 12 1.0
okra 13 50% 3 22 1.0
summer squash 10 45% 2 19 1.0
sage 4 26% 26 315 0.9
poppy seeds 1 17% 23 525 0.9
Chinese cabbage 14 54% 2 12 0.9
watercress 20 65% 2 11 0.9
chives 12 48% 4 30 0.9
broccoli 13 50% 5 35 0.9
edamame 8 41% 13 121 0.9
sauerkraut 6 39% 2 19 0.9
jalapeno peppers 4 37% 3 27 0.9
cloves 6 35% 35 274 0.9
cauliflower 11 50% 4 25 0.9
marjoram 4 31% 27 271 0.9
caraway seed 3 27% 28 333 0.8
thyme 5 34% 31 276 0.8
red peppers 6 40% 3 31 0.8
radishes 7 43% 2 16 0.8
celery 10 50% 3 18 0.8
portabella mushrooms 12 55% 5 29 0.8

eggs and dairy


food ND % insulinogenic insulin load (g/100g) calories/100g MCA
egg yolk 5 18% 12 275 1.2
whole egg 6 30% 10 143 1.1
cream -6 6% 5 340 1.0
sour cream -5 13% 6 198 0.9
limburger cheese -1 19% 15 327 0.9
cream cheese -5 11% 10 350 0.9
camembert -1 21% 16 300 0.8
feta cheese -1 22% 15 264 0.8
Swiss cheese -0 22% 22 393 0.8
butter -7 2% 3 718 0.8
blue cheese -1 21% 19 353 0.8
gruyere cheese -0 22% 23 413 0.8
edam cheese -1 23% 21 357 0.8
cheddar cheese -2 20% 20 410 0.8
brie -3 19% 16 334 0.8
Monterey cheese -2 20% 19 373 0.8
goat cheese -3 21% 14 264 0.8
muenster cheese -2 21% 19 368 0.8
gouda cheese -1 24% 21 356 0.8
Colby -2 21% 20 394 0.7
ricotta -2 27% 12 174 0.7

nuts, seeds and legumes


food ND % insulinogenic insulin load (g/100g) calories/100g MCA
sunflower seeds 3 15% 22 546 1.0
flax seed 0 11% 16 534 1.0
coconut milk -6 8% 5 230 1.0
sesame seeds -2 10% 17 631 0.9
brazil nuts -2 9% 16 659 0.9
coconut cream -7 8% 7 330 0.9
pumpkin seeds 1 19% 29 559 0.9
hazelnuts -2 10% 17 629 0.9
coconut meat -6 10% 9 354 0.8
walnuts -1 13% 22 619 0.8
almonds -1 15% 25 607 0.8
pine nuts -3 11% 21 673 0.8
almond butter -1 16% 26 614 0.8
pecans -5 6% 12 691 0.8
macadamia nuts -6 6% 12 718 0.7



food ND % insulinogenic insulin load (g/100g) calories/100g MCA
mackerel 0 14% 10 305 1.1
fish roe 15 47% 18 143 1.1
caviar 9 33% 23 264 1.1
cisco 5 29% 13 177 1.0
trout 13 45% 18 168 1.0
sardine 9 37% 19 208 1.0
sturgeon 14 49% 16 135 0.9
salmon 15 52% 20 156 0.9
anchovy 11 44% 22 210 0.9
herring 7 36% 19 217 0.9


food ND % insulinogenic insulin load (g/100g) calories/100g MCA
beef brains 3 22% 8 151 1.1
lamb brains 5 27% 10 154 1.1
sweetbread -3 12% 9 318 1.0
lamb liver 14 48% 20 168 1.0
turkey liver 13 47% 21 189 1.0
chicken liver 14 50% 20 172 0.9
liver sausage -4 13% 10 331 0.9
chicken liver pate 5 34% 17 201 0.9
lamb kidney 14 52% 15 112 0.9
veal liver 15 55% 26 192 0.8
liver pate -4 16% 13 319 0.8
lamb sweetbread 7 43% 15 144 0.8
beef kidney 11 52% 20 157 0.7

animal products


food ND % insulinogenic insulin load (g/100g) calories/100g MCA
bratwurst 0 16% 13 333 1.0
ground turkey 5 30% 19 258 0.9
bacon -4 11% 11 417 0.9
pork sausage 1 25% 13 217 0.9
salami -1 18% 17 378 0.9
pork ribs -1 18% 16 361 0.9
kielbasa -3 15% 12 325 0.9
turkey bacon -3 19% 11 226 0.8
pork sausage -2 20% 16 325 0.8
knackwurst -4 16% 12 307 0.8
roast pork 8 41% 20 199 0.8
bologna -7 11% 9 310 0.8
pepperoni -4 13% 16 504 0.8
beef sausage -3 18% 15 332 0.8
lamb rib -2 19% 17 361 0.8
duck -3 18% 15 337 0.8
pork ribs 6 39% 21 216 0.8
blood sausage -5 14% 13 379 0.8
pork loin 7 41% 19 193 0.8
frankfurter -5 17% 12 290 0.8
meatballs -3 19% 14 286 0.8
headcheese -5 20% 8 157 0.8
roast ham 6 41% 18 178 0.8
chorizo -3 17% 19 455 0.8
roast beef 5 38% 21 219 0.7
turkey -2 20% 21 414 0.7
chicken (leg with skin) 6 42% 18 184 0.7
T-bone steak -1 26% 19 294 0.7
ground beef 1 30% 18 248 0.7

other dietary approaches

The table below contains links to separate blog posts and printable .pdfs detailing optimal foods for a range of dietary approaches (sorted from most to least nutrient dense) that may be of interest depending on your situation and goals.   You can print them out to stick to your fridge or take on your next shopping expedition for some inspiration.

dietary approach printable .pdf
weight loss (insulin sensitive) download
autoimmune (nutrient dense) download
alkaline foods download
nutrient dense bulking download
nutrient dense (maintenance) download
weight loss (insulin resistant) download
autoimmune (diabetes friendly) download
zero carb download
diabetes and nutritional ketosis download
vegan (nutrient dense) download
vegan (diabetic friendly) download
therapeutic ketosis download
avoid download

If you’re not sure which approach is right for you and whether you are insulin resistant, this survey may help identify the optimal dietary approach for you.


Atkins versus the vegans

Dr Fung also noted that the Atkins approach often doesn’t work over the long term because things other than carbohydrates require insulin.

The food insulin index data demonstrates that a number of high protein foods such as steak, cheddar cheese, white fish and tuna cause a significant insulin response even though they contain minimal carbohydrate.


The irony of low carbers eating protein to avoid carbs to minimise insulin secretion although the insulin index data shows that protein foods cause a significant insulin effect has not been lost on the vegan community as shown in this thought provoking video below.

In response to this, Gary Taubes has acknowledged that protein does stimulate insulin, however has stated that

“the assumption has always been that this effect is small compared to that of carbohydrates, and that it is muted because protein takes considerably longer to digest.”

Is protein a significant issue an issue for people trying to control blood sugar and reduce the insulinogenic load of their food?

Does the fact that protein takes longer to digest mean that the insulin secreted in response to protein doesn’t matter?

Perhaps the food insulin index data can help us find the answer.

[next article…  how much insulin is required to cover protein?]

[this post is part of the insulin index series]

[Like what you’re reading?  Skip to the full story here.]