correlation of nutrient density with carbs, fat, protein, insulin load, saturated fat and cholesterol

  • Reducing the amount carbohydrates tends to improve nutrient density.
  • Nutrient density seems to peak at between 10 and 30% insulinogenic calories.
  • Monounsaturated fat from foods like olive oil, avocados and nuts seems to be correlated with higher overall nutrient density.
  • Higher protein seems to be correlated with improved nutrient density.
  • Cholesterol, polyunsaturated fat and saturated fat all seem to have negligible correlation with nutrient density.

Does everything we eat cause heart disease?

I recently came across an interesting observational study from 1999 where researchers compared different components of the diet with heart disease rates using data from more than one hundred thousand participants in the Nurses’ Health Study.[1]   The analysis was interesting but somewhat confusing.  It seems that pretty much everything we eat is correlated with heart disease to some degree, but it’s hard to tell which component of our diet is the worst culprit.


what do we know?

What we do know is that eating too much is not good for us.  Most of our western diseases are not due to malnutrition but rather over nutrition.  Our bodies just don’t deal well with too much energy on an ongoing basis.

We also know that most people struggle to regulate their appetite to achieve optimal weight or normal blood glucose levels.  Even when we resort to calorie counting we struggle to fight against the body’s desire to maintain a certain weight.  It’s hard to fight the drive to survive for too long.

I think the real challenge is to work out what we should eat that will help us to obtain adequate nutrients without requiring us to eat too much.  And then hopefully this will lead to regulation of our appetite and avoidance of overeating

The principle that the LCHF scene understands is that, for someone who is insulin resistant, eating foods that stimulate too much insulin will lead to excess energy storage and increased appetite. Fasting and / or a reduced insulin load dietary approach can help to reduce insulin levels and enable you to access more of your body fat stores for energy.

Once you’ve got your blood glucose and insulin levels under control and you still have more weight to lose you can focus more on maximising nutrient density of our food so we won’t keep on searching out more food to obtain more nutrients.

Then once blood glucose and insulin levels are normalised I think the next quest is to find what foods will maximise nutrition while enabling us to minimise energy intake.

Rather than worrying about protein, fat, carbohydrates we would be better served to focus on the foods that provide higher levels of nutrients that will leave us satisfied with less calories.  Rather than focussing on avoiding particular dietary components like saturated fat, cholesterol or carbohydrate might it be better to think in terms of what will maximise nutrient density.

The purpose of this article is to look at whether there is any relationship between nutrient density and any of the food parameters that we typically focus on.  Are there any parameters that correlate with nutrient density or are they all a waste of time?   To what degree does worrying about protein, fat, carbs, insulin load, saturated fat etc help us achieve a more nutrient dense diet that will lead to better health and avoid over nutrition?

I’m not saying that there is any causal relationship between these parameters.  The reality is that there are nutrient dense high carb foods and nutrient dense high fat foods.  However, I think that learning what is more useful and what is perhaps useless is still useful.

Nutrient density index

Listed below are the nutrients that contribute to the nutrient density score as detailed in the building a better nutrient density index article.  In this list below I have crossed out the nutrients that are fairly easy to obtain so we can focus on the ones that are harder to get in adequate quantities.


  1. Choline
  2. Thiamine
  3. Riboflavin
  4. Niacin
  5. Pantothenic acid
  6. Vitamin A
  7. Vitamin B12
  8. Vitamin B6
  9. Vitamin C
  10. Vitamin D
  11. Vitamin E
  12. Vitamin K


  1. Calcium
  2. Copper
  3. Iron
  4. Magnesium
  5. Manganese
  6. Phosphorus
  7. Potassium
  8. Selenium
  9. Sodium
  10. Zinc

amino acids

  1. Cysteine
  2. Isoleucine
  3. Leucine
  4. Lysine
  5. Phenylalanine
  6. Threonine
  7. Tryptophan
  8. Tyrosine
  9. Valine
  10. Methionine
  11. Histidine

fatty acids

  1. Docosahexaenoic acid (DHA) (22:6 n-3)
  2. Eicosapentaenoic acid (EPA) (20:5 n-3)
  3. Docosapentaenoic acid (DPA) (22:5 n-3)
  4. Alpha-linolenic acid (18:3 n-3)
  5. Arachidonic acid (20:4)
  6. Oleic acid (18:1)
  7. Lauric acid (12:0)
  8. Capric acid (10:0)
  9. Pentadecanoic acid (15:0)
  10. Margaric acid (17:0)

Direct or parabolic?

Engineers and scientists like to develop mathematical relationships to try to understand the relationships between different parameters.

The chart below shows an example of a direct relationship.  In this example we see the data from the Ancel Keys’ Seven Countries Study[2] that suggested that, based on data from these seven studies, that more fat in the diet the more deaths we have from heart disease.


Typically, things in nature are not direct relationships.  Often, some is better than none but too much is bad.  Think of it as the Goldilocks Zone.   Not too much, not too little, but just right.

It’s earlier to think in terms of good / bad, black or white, but many things in nature are not linear.  The chart below from Dr Ted Naiman demonstrates a parabolic relationship.  Often the optimal range lies somewhere between the extremes.


Proportion of insulinogenic calories

The relationship that is of most interest to me is the one between the proportion of insulinogenic calories versus nutrient density.

To recap, the insulin load is the amount of the food you eat that will require insulin and potentially raise your blood glucose levels.  The proportion of insulinogenic calories is the proportion of your diet that will require insulin (i.e. using amino acids for muscle protein synthesis and storing excess energy from non-fibre carbohydrates in our fat cells).

insulin load (g) = carbohydrates (g) – fibre (g) + 0.56 * protein (g)

The chart below shows that if we run a simple linear trend line (orange) then nutrient density declines with increasing insulinogenic calories.  However, if we run a polynomial trend line (red), we get a higher correlation.


Nutrient density appears to peak around 10 to 40% insulinogenic calories.  If you have less than 10% you end up with a very high fat low protein diet and you will have to pay extra attention to make sure you get adequate nutrition.  If you run with a very high insulinogenic diet you risk ending up with excessed processed carbohydrates that often have low nutritional value.

This is a broad generalisation.  The massive amount of scatter in this chart demonstrates that we can find nutrient dense foods and nutrient poor foods at either end of the scale.


Most people agree that highly processed carbohydrates are problematic and higher levels of fibre (or cellular carbohydrates[3]) is good.  The plot below shows the relationship between net carbohydrates (i.e. total carbohydrates minus fibre) and nutrient density.   We can see that with either a linear or polynomial trend line that nutrient density drops away as we increase carbohydrates.


On the left hand axis there are heaps of foods with zero carbs that have either high or low nutrient density.  What skews this analysis is the fact that foods with 100% carbohydrate and no fibre typically have a very low nutrient density.   It’s not that we should necessarily avoid all carbohydrates, but rather that we should avoid the processed ones and maximise the non-starchy veggies that have some fibre.

The story with total carbohydrates is similar.   The scatter in these charts is massive, but as a general rule it seems that foods with more carbohydrates have a lower nutrient density.  The take home message here seems to be that processed carbohydrates with minimal amounts of fibre are generally nutrient poor.



The plot below indicates that a higher levels of protein are associated with higher nutrient density.   Nutrient density peaks at around 40% protein.   [Keep in mind that a lot of the amino acids are not counted in the nutrient density score to make sure protein doesn’t dominate the scoring.]


Optimal protein levels are a contentious topic.  There is research out there that says that excess protein can be problematic from a longevity perspective.  Protein promotes growth, IGF-1, insulin and cell turnover which can theoretically compromise longevity.  At the same time there are studies that indicate that we need much more protein than the minimum DRI levels.[4]

Like excess anything, excess protein can be problematic.  In the end you need to eat enough protein to prevent loss of lean muscle.  If you’re trying to build lean muscle and actually working out, then higher levels of protein may be helpful to support muscle growth.  If you are trying to lose weight then higher levels of protein can be useful to increase satiety and prevent loss of lean muscle mass.

If you’re insulin resistant and trying to achieve ketosis, then higher levels of protein can be problematic due to the higher levels of insulin generated, particularly if you are not exercising.  The best approach here seems to be to eat as much protein as you can tolerate while maintaining excellent blood glucose levels and some ketones.  Over time you may be able to tolerate higher levels of protein as your insulin sensitivity improves and thus be able to improve the nutrient density of your diet.

The table below shows a range of minimum protein intakes that may be suitable for you depending on your height.  Lower protein levels may be appropriate if your are aiming for therapeutic ketosis while higher protein levels may be appropriate if you are looking to gain muscle and have a lower carbohydrate intake.


Lots of people will argue that you can’t go wrong with too much fat while others caution against excess fat.  The plot below shows the relationship between fat and nutrient density.  If you run a straight line trend line, then nutrient density increased with more fat.  This is at odds with what we have been told for the last four decades based on Ancel Keys’ Seven Countries study chart above.


The polynomial trend line (red ) suggests that nutrient density peaks at around 50 to 60% of calories from fat.   This level is probably a good starting point for most people.  People aiming for therapeutic ketosis to manage cancer or epilepsy can increase the fat levels or decrease fat if they are going for weight loss and have their blood glucose levels under control.  Personal preferences, culture and activity are all relevant considerations.

It’s interesting to see the change in thinking in even the highest levels of research and the USDA dietary guidelines committee.  In this Dr Meir Stampfer from Harvard, one of the authors on the epidemiological article above, talks about how their advice to minimise fat was “bad advice” due to:

  1. group think (all the previous advice had to been to minimise fat so it is hard to turn the ship around),
  2. basing advice on a market (i.e. LDL, cholesterol) rather than actual clinical end points, and
  3. the desire to be simple (i.e. they wanted people to avoid trans fats and saturated fats but just recommended avoiding all fat for simplicity which led to an increase in processed carbs).

saturated fat

The chart below shows that nutrient density seems to increase as we increase saturated fat although the correlation is very low.


Does this data mean you should eat as much saturated fat as you can?  No.

Does it suggest that saturated fat is not really an issue to be concerned about?  Maybe.


The plot below shows a plot of cholesterol versus nutrient density.  Again, it appears that nutrient density tends to be higher with higher levels of fat, including cholesterol.   At the same time cholesterol typically only forms a small proportion of most foods (i..e less than 0.75% of energy) and the correlation between nutrient density and cholesterol is very low.


In the past there has been an assumption that cholesterol in your food equates to cholesterol in your blood which leads to heart disease.   These days the wheels seem to be falling off this simplistic narrative.[5]  [6]

The recent 2015 Dietary Guidelines Committee dropped cholesterol as a ‘nutrient of concern’ due to the lack of evidence that there is a relationship between dietary cholesterol and cholesterol in the blood.[7]

Does this mean you should go out of your way to eat high cholesterol foods?  Probably not.

Does this suggest that you don’t need to worry too much about the amount of cholesterol in your diet?  It seems so, yes.

Monounsaturated fat

The chart below shows the relationship between mono unsaturated fat and nutrient density.


The simple linear trend line suggests that nutrient density improves with more mono unsaturated fat.  The highest correlation however is with the polynomial trend line with the peak nutrient density occurring at around 50 to 60% of calories from mono unsaturated fat.

This aligns with the ‘common knowledge’ that monounsaturated fats from olive oils, nuts and avocadoes should be considered to be “good fats” in terms of nutrient density.

polyunsaturated fat

The chart below shows the relationships between nutrient density and the amount of polyunsaturated fat.   It seems that more poly unsaturated fat from sources such as cold water fish, dairy, eggs and nuts will provide a higher level of nutrition, though the correlation is quite low.


which factors matter the most?

The table below shows the correlation of the different parameters analysed above with nutrient density, sorted from the highest correlation to lowest correlation.  I have noted the correlation coefficient as well as whether the highest correlation is achieved with a polynomial or linear trend line.  A polynomial relationship means that there might be an optimal level while a linear means that there is more likely to be a good / bad relationship.  While none of these relationships are ‘statistically significant’ the trend is interesting.

property corr relationship comment
net carbs 0.11 linear the lower the better
total carbs 0.095 linear the lower the better
% insulinogenic calories 0.083 polynomial 10 to 40% optimal
total fat 0.061 polynomial 40 to 70% appears optimal
protein 0.054 polynomial 40 to 50% appears optimal
monounsaturated fat 0.0094 polynomial monounsaturated fats = “good fat”
saturated fat 0.0020 polynomial very low correlation
cholesterol 0.0015 polynomial very low correlation
polyunsaturated fat 0.0012 polynomial very low correlation