Popular nutrition enthusiasts typically focus on the one magic factor that will fix your ills.
- blood type
- low carb
- low fat
“It worked for me. It’s the only way, and it will surely work for you!”
Meanwhile, both public health messaging and click bait-hungry media is awash with confusion.
Every week there is something new that we should be fearful of:
- saturated fat
- seed oils
Everyone seems to have their favourite component of food to demonise that aligns with their bias that has undoubtedly led to the obesity crisis that is spiralling out of control.
If you were prone to conspiracy theories, you could think that THEY want to keep you in a perpetual state of confusion, so you keep buying WHAT MAKES YOU FEEL GOOD rather than WHAT IS GOOD FOR YOU.
There is money to be made in keeping us moderately sick and feeding us food that is cheap to manufacture and hard to stop eating.
Then more recently, ethical and environmental considerations are starting to dominate the narrative, taking the focus off human nutrition (see Should You EAT Lancet).
With so many competing agendas it’s hard for the average person who wants to stay healthy to know what they should actually eat.
Satiety and nutrient density
When we obtain the vitamins, minerals and fatty acids from our food, we are less likely to over eat.
But what other parameters have a positive benefit?
The half a million days of MyFitnessPal data used in the satiety analysis only contains details of protein, fibre, carbs, fat and sodium. Unfortunately, we don’t have enough data to isolate the other elements of our food system that help us to gain control of our appetite.
So we thought it would be interesting to dig into the USDA database to see which nutrition parameters correlated with nutrient density, and hence which food properties are likely to help us avoid overeating due to the ‘hidden hunger‘ caused by a lack of nutrients.
Human metabolism and nutrition is complex and multi-factorial. There are many symbiotic moving parts that can’t be separated.
But if you were to compile a simple shortlist of food parameters that the general population should focus to manage your health, what would they be?
A quick primer in statistics
Before we get too deep into the data analysis, let me give you a quick primer on some of the basic statistical analysis used in this article.
Correlation is the degree to which two things are aligned.
You have probably heard the saying ‘correlation does not equal causation’. The chart of lemon imports and highway deaths shows that things can be correlated, but not causative.
We need to exercise some healthy scepticism about correlation. However, it can be useful when combined with other data as well as to disprove something or reality check a belief.
If two things are ‘positively correlated‘, they move together while things that are not correlated seem to have no influence on each other. Meanwhile, things that are negatively correlated move against each other (i.e. when one goes up, the other goes down like a seesaw).
If two things have a correlation of 1.0, they move in lockstep with each other. While, if two things have a correlation of 0, it is complete noise and there is no relationship. If we have a correlation of -1.0, they are inversely correlated.
- Energy density is negatively correlated with nutrient density. It seems that foods that have a higher energy density tend to be more processed have a lower nutrient density.
- Fibre and protein don’t necessarily cause nutrient density to increase. But , as a general rule, higher fibre and protein tend to be positively correlated with higher nutrient density.
Pearson correlation coefficient
The Pearson Correlation Coefficient is a measure of the degree of correlation between two things.
It is the measure of the distance of all the points from a straight line through a group of points on an X, Y plane. If all the points sat neatly on a straight line, we would have R2 = 1.
An R2 of 1.0 is extremely rare in real life where variations occur due to statistical measurement errors.
Understanding R2 enables us to understand which factors have the most significant impact on nutrient density, and conversely, the things that have very little bearing on nutrient density and will likely provide minimal benefit in terms of human nutrition.
Linear trendlines vs polynomial trendlines
A trendline is simple line of ‘best fit’ through the data.
With a linear trend line, we draw a straight line through the data, and we can get the R2 for that trendline.
Many people want to think in terms of good or bad, on or off. Carbs / sugar / fat / fructose / protein (or whatever) is either good or bad. It’s much simpler and easier to think in terms of black and white rather than optimal. However, black and white thinking often can gets us into trouble.
In the real world, linear relationships are rare. There is often a peak or an optimal value. You can have too little or too much of something.
Using a polynomial trendline (like the one shown below) enables us to see the point at which nutrient density peaks and then decreases once we exceed the optimal amount.
And what is a lever anyway?
In engineering, you use leverage to move something heavy with a small force at a longer distance. In finance you can borrow, or ‘use leverage’, to achieve a bigger impact with your available cash.
When it comes to nutrition we think it is valuable to know which actions you can take that will provide the most beneficial outcome.
We want to find the minimal effective dose of effort to get the maximum benefit from your investments in your health and nutrition to help you navigate the modern food environment.
The objective of this analysis is to identify the food parameters (or combination of parameters) that have the greatest positive impact that we can use as part of the Nutrient Optimiser algorithm to help fine tune your diet. Conversely, we can also identify the parameters that provide the least benefit.
Many approaches to nutrition rely on outdated (often religious) belief systems, ethical viewpoints or anecdotes. Unfortunately, these often get us into trouble when taken to extremes.
The data analysis
Foods with a higher nutrient density are those that have more of the nutrients that are harder to find in adequate quantities. The ‘harder to find’ nutrients included in this analysis are:
- vitamin E,
- vitamin D,
- pantothenic acid,
- choline, and
- omega 3.
Other nutrients are still important, but they are typically easier to find in our food system (based on the USDA nutrient database), so we less likely to be deficient in the nutrients not listed above.
Nutrient density is context specific based on food availability and your current food preferences. However, these are the twelve nutrients that are generally harder to find in the food system.
To help us to understand the relationship between nutrient density and various other parameters, the table below shows:
- the correlation between nutrient density and various nutritional properties (if the value is positive then it means that this property has beneficial and vice versa),
- the R2 value if we assumed there was a linear relationship, and
- the R2 value if we assume there is a polynomial relationship.
The table is sorted based on the polynomial R2 value (right-hand column) from highest to lowest). The parameters at the top of the table are more closely correlated with nutrient density (and, by inference, satiety, weight maintenance and improved metabolic health). Meanwhile, the parameters at the bottom of the table seem to have little bearing on the overall quality of our diet.
|parameter||correlation||linear R2||poly R2|
In the following sections, we’ll dig into each of the parameters separately in descending order, from most significant to least significant.
The chart below shows the relationship between fibre and nutrient density.
Foods with more fibre tend to have a higher nutrient density. With a Pearson Correlation Coefficient (R2) value of 0.42 we can see that, while it’s not the only factor, it is an important factor.
This observation aligns with the laboratory data from A Satiety Index of Common Foods (Holt et al, 1995) which showed that foods that contain more fibre were more satiating and thus harder to overconsume.
The satiety analysis of half a million days of food records from ten thousand people also suggests that foods with more fibre tended to produce greater satiety.
It’s not that fibre causes food to be more nutritious, but rather that foods with more fibre tend to contain more of the nutrients that tend to be harder to find.
While you can’t just dump a pile of fibre supplement into your diet and expect an immediate health benefit, foods that are less processed and in their natural form (without fibre removed) tend to contain more vitamins and minerals.
The chart below shows that foods with more protein tend to contain more of the harder-to find-vitamins and minerals. The polynomial trendline suggests that nutrient density peaks at around 70% calories from protein.
This observation also aligns with our previous analysis of the nutrient density of various dietary approaches, with a peak in protein at a fairly high 45% of calories.
While more protein seems to be better, it’s not practical for most people to consume very high amounts of protein due to the strong satiety effect. As we can see in the chart below from the satiety analysis, most people struggle to consume more than around 35% of their energy from protein. Our appetite craves energy as well and will go in search of energy from protein and carbs once it has enough protein.
Target protein intake levels are contentious and are dependent on a lot of factors. Minimum protein intake levels are typically set by nitrogen balance studies which tend to arrive at a much lower level.
To cut through the confusion, we have programmed Smart Macros into the Nutrient Optimiser to ensure you are getting enough protein to prevent excessive loss of lean mass.
- If your goal is weight loss, and you are losing more lean mass than body fat, the Nutrient Optimiser will increase your protein intake by 5g each week to ensure that you maximise retention of lean muscle mass.
- If you are a bodybuilder trying to build lean mass, the Smart Macros algorithm will tweak your protein intake to ensure you are gaining more muscle than fat.
Protein generally becomes a non-issue when people focus on getting the harder to find vitamins, minerals and essential fatty acids.
The chart below shows that foods with a higher energy density tend to have a lower nutrient density.
This observation aligns with the laboratory data from A Satiety Index of Common Foods (Holt et al, 1995). Foods with a lower energy density (e.g. grapes, oranges, porridge and fish) tend to be more satiating than foods with a higher energy density (e.g. cakes, doughnuts and peanuts).
Foods that contain omega 3 tend to have a higher nutrient density.
The chart below shows that lower levels of non-fibre carbs align with a higher nutrient density.
Once you have had adequate protein, the balance between obtaining fuel from fat vs carbs largely comes down to food preference, availability and demands to support explosive activity.
If you are trying to manage your blood sugars, the Nutrient Optimiser Smart Macros algorithm can help you to find your carbohydrate tolerance by titrating down your non-fibre carbohydrate limit by 5g per week until they stabilise.
If your blood sugars are stable, Smart Macros will progressively increase your carb limit to enable you to focus on maximising nutrient density and fueling performance.
Keto is quite popular at the moment, with many of people targeting foods with fewer carbs and more fat, and sometimes less protein.
The chart below shows the relationship between % insulinogenic calories and nutrient density, indicating that there is an optimal middle ground between the extremes.
Foods to the far right contain less protein and non-fibre carbohydrates (i.e. keto foods), while foods to the left contain more non-fibre carbohydrates.
People who come from a highly processed insulinogenic diet that has made them obese and diabetic will benefit by reducing processed carbs and not fearing fats. However, moving too far to the other extreme with a significant proportion of your diet consisting of refined fats and processed oils with minimal protein is not ideal in terms of nutrient density.
Omega 6 PUFAs
The chart below shows that foods with more omega 6 “vegetable oils” or PUFAs tend to have a lower nutrient density.
Foods with very high levels of omega 6 include corn oil, soy oil, and dressings and mayonnaise that contain these oils.
Since we discovered how to extract oil out of soybeans and rapeseed using chemical solvents, the use of omega 6 “vegetable oils” in our food supply has exploded!
Cheap and hyperpalatable, vegetable oils are perfect for food manufacturers in terms of sales, and perhaps our personal budget, but not our health and our waistline.
PUFAs can also become oxidised at high temperatures and have a number of inherent health risks. It is also very hard for most people to achieve an optimal omega 3:omega 6 ratio with a significant amount of polyunsaturated fats in their diet.
As a general rule, most people would benefit from eating more seafood and fewer products that have seed oils on the ingredients list to bring their omega 3 to omega 6 ratio back towards optimal.
As shown by the polynomial trend line in the chart below, foods with moderate levels of starch seem to be the most problematic in terms of nutrient density. Foods with a very high proportion of energy from starch may not be a big concern (in the absence of added fat).
The Holt (1995) laboratory satiety data also indicated that high starch foods were harder to overeat than moderate starch foods.
Very high starch foods such as plain potatoes, rice and plain pasta are hard to overeat, but very few people eat potatoes and rice without some added fat which makes it much easier to overconsume.
This observation is consistent with the analysis of half a million food diaries where we seem to maximise calorie intake with moderate levels of starch, while there seems to be a slight decline as starch levels increase to very high levels when we remove fat from the diet.
Saturated fat has a low correlation with nutrient density.
However, while saturated fat is probably not the primary cause of cardiovascular disease, insulin resistance and obesity that some people make it out to be, it is not a nutrient-dense super food either.
Foods that are mostly saturated fat just don’t contain a lot of vitamins and minerals. But at the same time, unless you are trying to lose weight aggressively you need to get your energy from somewhere.
Over the past hundred years consumption of saturated fat has gone down as a percentage of our total intake.
The availability of saturated fat in our food system has increased with a rise in overall energy intake, although not as much as monounsaturated fat and polyunsaturated fat which have experienced massive growth over the past hundred years.
Some people portray fructose (aka fruit sugar) as the ultimate boogeyman. However, the chart below shows that moderate levels of fructose align with a higher nutrient density.
Foods with a moderate amount of fructose include lettuce, cauliflower, cucumber, celery tomato, squash, cabbage, beans, radish.
Fruits like apples and oranges provide even more fructose and don’t offer a lot of nutrients, but they tend to be hard to overeat in their natural form due to their lower energy density (i.e. more water and fibre).
At the far right-hand end of the chart, separating energy from nutrients is never a great idea. Foods sweetened with high fructose corn syrup provide little nutritive value, other than a sharp burst of energy, which most of us don’t need very often.
Similar to fructose, foods that contain a moderate amount of glucose (such as mushrooms, asparagus, pickles and lettuce) do not constitute a significant concern.
However, foods that contain high levels of glucose such as honey, fruit juices and dried fruits are not a great idea if your goal is weight loss because they provide an energy dense hit without a lot of nutrients.
Higher-fat foods have a lower nutrient density. However, overall, total fat does not have as strong a correlation with nutrient density compared to other factors.
Note: We need some essential fatty acids to thrive. However, the actual quantity of essential fatty acids is not a lot. We can achieve a robust nutrient profile with a minimum of 0.4 g/kg LBM dietary fat or around 10% of maintenance calories. Anything above this is mostly used as fuel or stored if not used.
Monounsaturated fatty acids
High levels of monounsaturated fatty acids do not correspond with high levels of vitamins and minerals. Foods high in MUFAs include avocado oil, olive oil, macadamia nuts, hazelnuts and almond oil.
The way some people talk about sugar, you would think it was universally bad.
However, it seems that some level of sugar could be beneficial in terms of nutrient density.
Sugar found in whole foods such as fruits and vegetable is not going to be a concern for most people. However, added sugars (i.e. any ingredient ending in “ose”) should ideally be treated as a recreational drug.
It’s interesting to note that the amount of added sugar in the food system has decreased over the past two decades since the introduction of artificial sweeteners while obesity rates have continued to rise.
Thinking in terms of total carbohydrates does not appear to help us substantially improve nutrient density.
As noted above, some glucose and fructose from vegetables are correlated with higher levels of vitamins and minerals. Fibre (which is part of the carbohydrate family) also tends to correspond to higher levels of micronutrients. Talking about carbohydrates as a whole doesn’t seem to be particularly useful.
So armed with this knowledge of the positive and negative factors that affect nutrient density, perhaps we can combine them to improve their predictive power even further?
Non-fibre carbohydrates + fat
If there was one thing to avoid it would be foods that are a combination of non-fibre carbohydrates and fat.
Protein + fibre
Conversely, prioritising both protein and fibre will help to improve nutrient density and satiety.
As shown in the table below, these two combined factors have greater predictive power and correlation with nutrient density.
|parameter||correlation||linear R2||poly R2|
If I ruled the world
When it comes to identifying nutritious foods, some food parameters are more useful than others.
It seems that there is limited benefit worrying about our intake of:
- saturated fat,
- total fat,
- monounsaturated fat,
- total carbohydrate, or
If I ruled the nutrition world, I would dedicate the limited bandwidth available towards the messaging that is likely to give the most useful outcome that would lead people to improve the satiety and nutritional content of their diet, as summarised in the table below.
|Fibre||Prioritise whole foods that contain fibre over highly processed grains and oils (fibre supplements don’t count).|
|Protein||Prioritise higher protein foods. Your appetite will drive you to get enough protein, but you will find it hard to over consume.|
|Energy density||Prioritise unprocessed foods which will have a lower energy density over refined grains and oils which are easier to over consume.|
|Omega 3||Prioritise seafood. It contains omega 3 fatty acids as well as a wide range of other beneficial nutrients.|
|protein+fibre||A diet with more protein and fibre will tend to be more nutritious and satiating.|
|non-fibre carbs+fat||Avoid foods and meals that combine non-fibre carbs and fat.|
The exact food lists that empowered thousands of Nutrient Optimisers to gain control over their health without relying on fad diets.
|Nutrition Approach||most popular food lists|
|maximum nutrient density||Get it now for free|
|weight loss (satiety & nutrient density)||Get it now for free|
|blood sugar & fat loss||Get it now for free|
|blood sugar & diabetes||Get it now for free|
|nutrient-dense maintenance||Get it now for free|
|maximum satiety||Get it now for free|
|ketogenic||Get it now for free|
|athlete & bulking||Get it now for free|
|nutrient-dense carnivore||Get it now for free|
|nutrient-dense plant-based||Get it now for free|
|carnivore (maintenance)||Get it now for free|
|plant based (maintenance)||coming…|