The project, led by Professor Dariush Mozaffarian, aims to rank foods based on their healthfulness for front of package labelling.
Twitter lit up with dismay over the many nonsensical results.
Confusion and conflicts of interest abound in the field of nutrition. So, creating an unbiased quantitative ranking system that food manufacturers can’t game and manipulate is critical to combat the exploding diabesity epidemic.
When so many things that we do these days are optimised by quantitative and complex algorithms (Google, Instagram, Facebook and TikTok), doesn’t it make sense that we should have a to thrive?
Imagine if there was a “Google for nutrition” that gave you a shortlist of more optimal food and meals that align with your context, goals and preferences?
With the explosion of voice search, it won’t be too long before people are coming home and saying ‘Google/Alexa, what should I eat tonight/this week?’ and Amazon or Uber Eats will arrive on their doorstep with dinner or the groceries for the coming week without giving it further any thought.
Unfortunately, most ranking systems are built to advertise to YOU more precisely and end with YOU as the product. Facebook, Instagram and Google have quickly joined the ranks of the wealthiest companies because they do this so effectively.
Similarly, if the food industry finances a food ranking system, it’s likely to guide you to buy more of the products that the food industry wants to sell you.
With this in mind, I’ve spent a LOT of time over the past five years analysing and working to create a robust data-driven system to identify a shortlist of optimal foods and meals for different goals.
Food Compass is being proposed to be used for food labelling to help people make better choices in the aisles of your shopping centre, so this is important.
To some, diving into the data to understand how everything is interconnected might sound a bit drab. But it’s REALLY fun (at least for me) and REALLY important (for everyone, including you).
Ready for a data-driven adventure?
- The ranking factors
- Nutrient ratios
- Food-based ingredients
- Specific lipids
- Which nutrients matter in a food ranking system?
- The limitations of data
- The nutrient fingerprint
- The Optimal Nutrient Intake Targets
- Comparison of ONI score vs Food Compass
- How Can I Calculate My Nutrient Intake?
- Level Up Your Nutrient Density
To set the scene, the table below shows the domains used by the Food Compass System.
Each of the following domains has various subcategories:
- Nutrient ratios,
- Food-based ingredients,
- Specific lipids,
- Fibre and protein, and
In all, there are 35 separate ranking factors.
On the surface, these all sound logical. But will this complex system effectively help you identify foods that contain the essential nutrients you require to thrive and keep you satisfied?
Let’s take a look at the details and the data.
Let’s start with the big guns – protein and fibre.
We are privileged to have access to large and unique datasets that have enabled us to understand how macronutrients and micronutrients affect how much we eat.
We know from our previous analyses that both protein and fibre have a positive impact on satiety.
As shown in the chart below of protein % vs calorie intake, data taken from the analysis of our 125,761 days of data from 32,192 people who have used Nutrient Optimiser over the past four years shows that protein percentage has a massive impact on the amount of food you tend to eat.
People who consume a diet with a very low percentage of protein tend to eat twice as many calories as people who eat a very high percentage of protein.
Our appetite is continually balancing our requirements for nutrients (including protein) and energy. It’s critical to get adequate nutrients without too much energy.
The frequency distribution chart below shows that we tend to gravitate to an intake somewhere between high and low protein extremes.
The nutrient-calorie satiety response curves (like the one shown for protein) are created by dividing the data into ‘buckets’ based on a narrow range of nutrient intake (e.g. 10 – 15% protein) and calculating the average daily calories consumed by those people. With such a large dataset for both macros and micros, we can gain some powerful insights.
All our analysis of various data sets aligns with Raubenheimer and Simpson’s work around protein leverage. As documented in their research over decades, protein leverage has been consistently observed in a wide range of organisms, from slime to insects, to animals in the wild and humans.
ALL organisms eat until we get enough protein. So if we only have access to low protein foods, we eat more. But if we only have access to higher protein % foods, we naturally consume fewer calories.
Before we go on, I want to address one common point of confusion: Protein leverage is not about simply eating MORE protein (i.e. in grams).
Simply eating more protein leads to consuming more calories. You can’t eat more doughnuts or butter (which have a low % protein) to get your protein.
You have to change what you eat. In practice, this requires a reduction in easily accessible energy from fat and carbs and a modest increase in protein in grams (as shown in the chart below from Protein Leverage: Theoretical Foundations and Ten Points of Clarification by Raubenheimer and Simpson).
The chart below the relationship between protein and non-protein energy (from fat and carbs) in percentage terms.
Fibre also has a positive impact, but it is not as large as protein. When we overlay protein and fibre together, we see that the potential impact fibre alone is relatively small.
When I spoke with Professors Raubenheimer and Simpson recently about my Nutrient Leverage Hypothesis, they asked me if I’d run a multivariate analysis on the data to understand if each of the micronutrients had an independent effect on satiety.
So that’s what I’ve been doing intensely for the past few weeks, with a fresh dump of data from Nutrient Optimiser, to understand the significance of the various quantifiable parameters of the food we eat.
But what is multivariate analysis?
I’m glad you asked.
A multivariate analysis helps us understand the relative impact and statistical significance of various parameters in a complex system (like our food matrix).
The single variate analysis (with protein alone), shown in the table below, demonstrates that:
- the relationship between protein % and calorie has a P-value of 0, mean that the relationship between protein % and calorie intake is highly statistically significant (i.e. not due to chance), and
- moving from 20% to 44% protein aligns with a 24% reduction in calorie intake.
But if we wanted to understand how both protein AND fiber are related to our total calorie intake, we could use a multivariate analysis as shown in the next table.
From this, we see that:
- increasing protein from 20% to 44% corresponds with a 23% reduction in calories, while
- moving from 3% to 9% fibre aligns with a smaller 4% reduction in calories.
Prioritising both protein % and fibre align with a lower calorie intake, but protein % is six times more effective than fibre.
Food compass weighting for fibre and protein
The good news: Food Compass has included fibre and protein in their ranking factors.
The bad news: The weighting for these important parameters is very low.
The chart below shows the relative weighting given to each of the domains of ranking factors used by Food Compass. The little brown slice in the top left represents protein and fibre combined. Only 6.7% of the weighting in their system was given to protein and fibre!
The fact that some of the highest-ranking foods contain negligible protein (e.g. celery juice, cheerios, grapefruit, avocado & chocolate almond milk) suggests that they might have dropped the ball in the design of Food Compass if their goal was to empower people to eat less and reverse the diabesity epidemic.
Fibre and protein certainly deserve to be included. However, Food Compass appears to have massively underweighted these important factors.
Protein tends to be more expensive to produce. Profit margins are lower. Hence, a system that guides consumers away from higher protein is good for the bottom line of food manufacturers.
Who knows if this was intentional or just poor system design in the Food Compass System. But it’s not going to help you identify foods that will satisfy your hunger or satisfy your body’s primary nutritional requirements.
Fun fact: The word protein is derived from the Greek word (and firstborn son of Poseidon) Proteus, meaning “of first importance”.
A complex system with many moving parts is not necessarily better than a simpler system that works. You can’t just throw a laundry list of your favourite indicators in a box, shake it up and hope for the best.
In modelling or forecasting (my primary domain as an engineer), complex black box systems often lead to garbage outputs, especially if the outputs can’t be calibrated with real-world data (e.g. how much people eat in different scenarios every day in the real world).
Before we go any further, I want to give you the punchline of our satiety analysis and this whole article:
Getting adequate protein without excess energy is the first fundamental step in optimising your nutrition and increasing satiety. If you read no further, this is the simple take-home message.
But if you want to learn more about the icing and learn to leverage nutrient density in your favour, then read on.
Our analysis also shows that minerals, to varying degrees, satisfy our cravings and lead to greater satiety (i.e. we eat less without having to exert unsustainable conscious willpower).
The charts below show the satiety response curves for the minerals considered by Food Compass. In each chart, you will notice that getting more of each of the minerals per calorie tends to align with a lower intake of calories.
However, you should also note that this is not a linear relationship. Once we get enough of these nutrients, our craving tends to decrease for that nutrient. Like a self-guided nutrient seeking missile, our appetite then sends us in search of foods that contain the other nutrients we need to thrive.
To understand the degree to which nutrients are beneficial in the food system, I have run a multivariate analysis of the essential minerals (together with protein, which we have already established, as seen above, has a strong relationship with satiety).
The analysis shows that a number of the minerals have a high degree of statistical significance. Protein accounted for a 23% reduction in calories when we only considered fibre, whereas, if we add in the minerals, protein accounts for only 17% of the reduction in calories, with 11% of the satiety reduction in calories attributed to the minerals.
It’s important to note that we will have different cravings for nutrients in different situations. If you are already getting plenty of a particular nutrient relative to your requirements for it, our craving for that nutrient will be lower (and vice versa).
When considered as a whole, this dataset shows a statistically significant satiety response to potassium, calcium, selenium, and sodium. However, in another group of people (or a smaller subset of this data), the statistically significant minerals may differ (i.e. context matters).
I have done a range of tests (which I’ll be sharing later) to understand the specific nutrients that have a stronger correlation with satiety for various dietary approaches (e.g. low protein, low fat and low carbs etc.).
But to summarise, for now, if the goal is to empower people to satiate their appetite and eat less, and thus reverse the diabesity epidemic, it appears that, in addition to protein, minerals are a worthy inclusion.
Next, we’ll look at vitamins.
Shown below are the nutrient-satiety response curves for the vitamins considered by Food Compass. Similar to protein and minerals, vitamins tend to have a positive effect on satiety, but only to a point. When we pass beyond the amount that can be obtained from whole food, the satiety benefit dissipates (and often rebounds).
Just because your Cheerios are fortified with a smattering of synthetic vitamins in an attempt to replace the nutrients lost in processing (and make your pee bright yellow), it doesn’t mean they’re more satisfying or necessarily healthier for you overall.
The table below shows the multivariate analysis for vitamins with protein. We see that, in addition to protein, some of the vitamins correspond with a modest reduction in calories.
While still significant, the reduction from vitamins is not as large as the effect we see from minerals. For instance, while consuming foods that contain more potassium leads to a 4% reduction in calories, consuming foods that contain more vitamin A leads to a smaller 2.1% reduction in calories.
Unfortunately, this multivariate analysis with vitamins is likely confounded somewhat by the fortification of isolated synthetic vitamins in otherwise nutrient-poor processed foods. The multivariate analysis assumes there is a linear relationship. So the analysis will be skewed by supplementation and fortification. If we could ever get a dataset with no fortification, it’s likely the satiety response to vitamins would be larger.
It’s interesting to note here that the availability of Vitamin A, the vitamin with the strongest correlation with how much we eat, has dropped significantly in our food system since the 1977 Dietary Guidelines for Americans (published by the US Department of AGRICULTURE) encouraged us to reduce animal-based foods (due to apparent concerns about saturated fat and cholesterol) and prioritise the products of industrial agriculture.
Next, we’ll look at the nutrient ratios used by Food Compass, i.e.:
- Potassium:sodium ratio,
- Fibre:carbohydrate ratio, and
- Unsaturated:saturated fat ratio.
As noted above, our nutrient-satiety analysis shows that sodium and potassium have a beneficial impact on satiety. We eat less of the foods that naturally contain more of these minerals per calorie.
As shown in the charts below, the availability of both of these critical minerals has also declined significantly in recent times as our soils have become depleted of these vital nutrients. Moreover, our previous analysis indicates that the growth in obesity is correlated with the decline in both of these minerals (see The biggest trends in nutrition for more details). The decline in sodium actually has the strongest correlation of any of the nutrients with the increase in obesity rates.
But there is also an upper limit to the benefits of these nutrients. That is, you won’t crave more of a nutrient you are already getting plenty of. Based on the satiety analysis, we have established an Optimal Nutrient Intake (ONI) of:
- 5 g sodium per 2000 calories, and
- 6 g potassium per 2000 calories.
This limit has been set based:
- The 85th percentile intake from our data from Optimisers, and
- The intake at which the effect starts to dissipate (based on the nutrient-satiety charts).
We wanted to ensure that the ONIs were achievable with whole food. It’s a stretch target, but it is achievable. However, there is no additional satiety benefit in reaching the 85th percentile. So we have set the ONI based on the point at which there is no additional satiety benefit.
The Optimal Nutrient Intake targets serve as both an aspirational goal and a limit. Even though processed food manufacturers hijack our cravings for salt, you don’t get a higher score if you are getting more than the Optimal Nutrient Intake for sodium (we’ll discuss this further later).
There is plenty of controversy around sodium. Most recommend trying to minimise it. However, sodium is not necessarily a nutrient to be minimised or avoided at all costs. The recent PURE study suggests results suggest that around 4 g of sodium per day aligns with the lowest risk of death from any cause.
However, we do need to keep it in balance with potassium. We used to believe that sodium drove hypertension (i.e. high blood pressure). However, we now understand that elevated blood pressure has more to do with too little potassium than too much sodium (ref, ref, ref, ref). For more details, check out the article How many grams of sodium do you need per day?
So the potassium:sodium ratio is a reasonable marker to consider for inclusion, especially given that most people don’t get enough potassium. The chart below shows the distribution of potassium:sodium ratio in our 125,761 days of data. The median potassium:sodium ratio of our Optimisers using Nutrient Optimiser is 1.2 (i.e. on average, they are getting slightly more potassium than sodium).
The nutrient-satiety chart below shows that people getting more potassium than sodium tend to consume fewer calories. But this benefit tapers off at around six times as much potassium vs sodium.
However, the correlation analysis shows that the potassium:sodium ratio has a fairly low correlation with calorie intake relative to potassium and sodium alone.
The following table shows the multivariate analysis with the nutrient ratios. Again, it appears that the potassium:sodium ratio isn’t a particularly useful addition if we have already considered potassium and sodium separately (with a limit on how much sodium is deemed to be beneficial).
Next, we come to the fibre:carb ratio, which the Food Compass System has used to measure refining and processing.
The plot of carbohydrates vs calories shows that a greater intake of carbohydrates generally aligns with eating more.
But it’s not a simple linear relationship.
- Zero carb is not better than low carb.
- Low carb (approx 20% energy from carbohydrate) tends to align with the lowest overall energy intake.
- The combination of fat and carbs (together with low protein) tends to correlate with the highest energy intake. This is the signature of high-profit margin, low satiety, hyper palpable junk food.
- At the right-hand side of the chart, we see that a high carb+low fat diet corresponds to a slightly lower calorie intake than a mixture of fat and carbs.
Looking at the satiety response curve for fibre:carb ratio, we see that, up to around 0.3, there is a beneficial relationship in consuming more fibrous carbohydrates. While this is a valuable marker of minimally processed whole food, you’re unlikely to get the same benefit from just adding a fibre supplement to an ultra-processed diet.
When we look at the multivariate analysis, we see that protein % still has the highest positive impact on calories (i.e. 19% reduction), followed by potassium (4.9%) and fibre:carb ratio (1.9%).
The fibre:carb ratio does provide a better satiety response than fibre %. Hence, the statistical analysis indicates that the fibre:carb ratio is worth retaining.
Unsaturated:Unsaturated fat ratio
Things start to get interesting when we look at the different fat fractions.
Fat is an easily accessible energy source, so it shouldn’t be surprising to see that more energy from fat corresponds with a higher energy intake. But this is not a linear relationship. Similar to carbohydrates, lower is not necessarily better. There is no benefit in terms of satiety from dropping fat calories below 40%.
Saturated fat, up until about 30% of calories, aligns with a higher calorie intake. However, intake tends to decrease beyond this point. This is likely because foods that contain saturated fat, as a component of the ingredients, simply provide more highly palatable energy (e.g. a hamburger and chips fried in lard). However, saturated fat in whole foods containing more protein is not as much of a concern (e.g. steak and eggs).
Meanwhile, a greater intake of monounsaturated fat aligns with a greater intake of calories in a fairly linear manner.
Processed plant oils (which are mainly monounsaturated fat) have become ubiquitous in our food system as cheap ingredients to provide more energy in our food. The largest change in energy in our food system has come from ‘added fats and oils’ followed closely by refined flours and cereals.
Unfortunately, Food Compass will serve to reinforce and perpetuate the guidelines by supporting the industrial food industry’s use of the ingredients that have fueled the obesity epidemic while demonising the ones that have remained unchanged.
Data from the USDA Economic Research Service shows that while sources of saturated fat have gone relatively unchanged over the past 50 years, there has been a massive increase in ‘salad and cooking oils’ (i.e. hydrogenated plant fats).
Note: ‘Salad and cooling oils’ = soy, canola and corn oil.
Our use of plant fats has become ubiquitous across the world as cheap calories to fuel a growing population (both in number and size).
Since we worked out how to extract oil from industrial crops (e.g. soybean, rapeseed, canola etc.) in 1908, through the process of hydrogenation, the fat in our food system has increased by more than six hundred calories per person per day (as shown in the chart below produced using data from the USDA Economic Research Service)! It’s important to note here that less than 100 calories of this energy from fat has come from saturated fat, while the remainder has come from both monounsaturated and polyunsaturated fat (mainly from plant oils).
Interestingly, we don’t see a significant signal from polyunsaturated fat (PUFA). Once we get beyond 3% energy, more PUFA does not necessarily lead to a greater intake of calories (at least based on this dataset). PUFA also tends to be a relatively minor contribution to fat calories compared to monounsaturated and saturated fat.
The chart below shows the satiety response curve for saturated fat as a proportion of total fat. If more of your fat comes from saturated fat, you will tend to consume more calories, at least up to about 40%. However, if most of your fat is saturated, you will likely consume a little less than a more even mixture of saturated and unsaturated fat.
When we look at the multivariate analysis, we can see that:
- consuming more polyunsaturated fat and saturated fat has a small positive impact on satiety, while
- saturated:unsaturated fat ratio and monounsaturated fat (%) have a very low statistical significance (i.e. low P-value) and do not have a large impact on our overall intake.
While saturated fat is not a superfood to be prioritised, there is no benefit in demonising whole foods that contain it. The use of saturated fat as a negative ranking factor or plant oils as a positive ranking factor will prioritise processed foods over whole food animal sources for no apparent benefit, at least in terms of satiety.
So, in summary, for nutrient ratios, we see that:
- Both potassium and sodium have a positive effect on satiety,
- Potassium:sodium ratio also has a positive effect. But once we limit the score awarded to sodium (i.e. above 5 g per 2000 calories), there is no added benefit from considering the potassium:sodium ratio).
- Fibre:carb ratio does have a beneficial positive correlation with satiety and is a useful differentiator of refined vs unprocessed carbs.
- Once protein % is factored in, there is no benefit considering saturated fat as a negative ranking factor.
Food Compass has used several ranking factors based on the ingredients they contain. I’ll offer some thoughts on each of these briefly below.
Fruits are a great energy source and are hard to overeat by themselves, but they don’t tend to contain much in the way of nutrients per calorie.
For example, watermelon, which ranks highly in the Food Compass system, provides a small burst of vitamin A and C but negligible amounts of the other essential nutrients.
Fruit is also not ideal if your blood sugars are not great. Large swings in blood glucose tend to lead to increased appetite when your blood sugar comes crashing down.
In spite of the deficiencies, there are numerous fruits that get a score of 100 in the Food Compass ranking system.
Fruit and vegetable juices (which are absorbed very quickly and hence provide little in the way of satiety) also rank really well according to Food Compass.
Non-starchy veggies rank highly in terms of nutrients per calorie. So there is no real benefit in counting it as a category.
As shown in the example below for soybeans, beans and legumes are a great source of energy (and possibly some protein if you don’t eat meat, dairy or seafood). However, I don’t believe giving a bonus for beans and legume is necessary, in addition to nutrients per calorie.
Whole grains are more nutritious than refined grains, which have the nutrients stripped out in processing to make them more palatable. In spite of this being used as a ranking factor, even fortified cereals like Cheerios scored incredibly well?!?!
Nuts and seeds are a great source of energy (if you need it). But most people find them easy to overeat, especially if they believe they are a “superfood”. It seems bizarre that, with some extra weighting on the nuts category, chocolate covered almonds made it to the top with a score of 78! I would love it if chocolate coated almonds were a high satiety superfood. But the reality is I can eat them all day and never feel full.
Seafood tends to rank highly in terms of nutrient density. It is a great source of protein, omega 3 and a range of vitamins and minerals. But, it doesn’t need any additional help from a specific ranking factor.
Perhaps they needed to add a new factor to reverse the unnecessary anti-saturated fat and cholesterol ranking factors that penalise dairy?
As discussed above, counting “plant oils” (e.g. monounsaturated fat extracted from industrial crops extracted via hydrogenation) as a positive ranking factor is unlikely to add significant value to a food ranking system.
In fact, I think it is ludicrous!
If you wanted to name a single smoking gun of the diabesity epidemic, you could reasonably point the finger at the 600 or so extra calories of fat (mainly from from plant oils made possible by hydrogenation) that have been added to our food system over the past century.
To reiterate, it’s the change in plant oils (aka ‘salad and cooking oils’ or ‘vegetable oils’) that has tracked closely with the rise in obesity. Meanwhile, our use of butter, lard and dairy fats has remained stable.
While saturated fat and plant oils behave slightly differently in our body, it’s not necessarily that one is better than the other. I think the main issue is that plant oils (produced by hydrogenation and fueled by synthetic fertilisers) are just a cheap source of energy to add to the food system, typically separated from protein and nutrients.
Red meat is an emotive subject for many, tied to beliefs around sustainability and ethical beliefs. But, then, at the other extreme, we have the carnivores eating nothing but steak and thriving.
But, red meat can be a cost-effective source of protein for many people who are not getting enough, particularly in developing countries, where people can’t afford to eat a vegan diet AND buy supplements to avoid deficiencies.
Processed meat is certainly not ideal but tends to rank lower than whole foods when viewed through the lens of nutrient density.
Overall, these qualitative judgements are unlikely to add additional value beyond the nutrient density. They simply reinforce the ‘conventional wisdom’ that has been built up over the past 50 years.
Weighting these belief-based nutritional factors above parameters that we have statistically significant data for (e.g. protein % and fibre) will dilute the usefulness of the recommendations.
The specific lipids considered by Food Compass are:
- Dietary cholesterol,
- Alpha-linolenic acid (ALA),
- Omega 3
- Medium-chain fatty acids (MCT), and
- Trans Fats.
Our analysis shows that foods that contain more omega 3 tend to have a positive impact on satiety. Seafood is high in omega 3, as well as protein and a range of beneficial vitamins and minerals.
You may be surprised to see that a higher intake of dietary cholesterol positively impacts satiety.
While not an essential nutrient, some amount of cholesterol has beneficial effects in the body (e.g. in the production of hormones). As shown in the chart below, our cholesterol intake has declined since the explosion of widespread industrial farming practices and the first Dietary Guidelines for Americans (published by the US Department of AGRICULTURE) that recommended that we consume less fat and avoid cholesterol. Sadly, since then, obesity has exploded.
The satiety response to trans fats is interesting. None seems to be better than some, but more is not worse. But, if you focus on nutrient-dense, minimally processed foods, you won’t be getting a large amount of dangerous trans fats. However, trans fats from whole foods sources aren’t necessarily a concern.
While potentially beneficial, data on alpha-linolenic acid and medium-chain fatty acids is generally hard to come by in food databases. Hence, if it is not regularly measured in food, it’s probably not a useful ranking factor for widespread use.
The multivariate analysis results below show that foods that contain more cholesterol provide greater satiety and align with a 4% reduction in calorie intake. At the bottom of the table, we see that trans fats, omega 3 and saturated fat don’t have a statistically significant impact.
Food Compass has also considered the following additives:
- Added sugar,
- Artificial sweeteners, flavours or colours,
- Partially hydrogenated oils,
- Interesterified hydrogenated oils, and
- High fructose corn syrup.
Additives are typically not ideal. However, they are seldom contained in nutrient-dense whole foods.
While we can fortify foods with a smattering of vitamins and minerals to add to the front of the box, foods that are nutritious across the spectrum are low in additives like these.
Again, these subjective parameters are unlikely to provide significant additional value in a ranking system.
Food Compass has also considered processing as an additional ranking factor.
Quantifying foods based on the level of processing may be useful on its own. However, again, it is unlikely to add additional value when we have first quantified foods that naturally contain a broad spectrum of nutrients, particularly protein.
Ultra-processed foods are typically produced using large-scale industrial agriculture, fueled by fossil fuel-based fertilisers that lead the crops to grow quickly but accumulate minimal nutrients from the depleted soils that are grown year after year. Nutrient density alone tends to eliminate these quickly.
Fermentation is generally a good thing. In our Nutritional Optimisation Masterclass, many people find they need sauerkraut to get their vitamin C or kombucha to get their calcium. However, it is not necessarily useful as an additional ranking factor.
Again, frying is irrelevant once we consider nutrient density. Frying simply accumulates lots of added fat which will dilute the nutrient density score of the food. So fried foods automatically fall to the bottom of the ranking tables.
Food Compass has also included phytochemicals (i.e. flavonoids and carotenoids). These are non-essential but potentially beneficial compounds contained in our food.
Unfortunately, these are not commonly measured and unavailable in most nutrition databases. As per Joel Furhman’s ANDI, a system that relies heavily on these parameters tends to bias heavily towards plant-based foods for which this data has been measured.
There is plenty of disagreement in various dietary camps about whether these non-essential plant compounds are beneficial or detrimental. Unfortunately, quantitative data is not yet available to understand the complex interactions of these non-essential nutrients.
As a general rule, if you are getting plenty of essential nutrients from whole food sources, you will also be getting plenty of the non-essential nutrients such as phytochemicals as a bonus.
So far, in this article, we have evaluated a range of nutrient parameters, some of which appear to have a statistically significant impact on satiety.
To help us understand which nutrients move the satiety needle, the table below shows the multivariate analysis of the micronutrients that have the most significant beneficial impact on satiety.
Protein % is by far the most significant factor. However, in this dataset, the nutrients that also exert a statistically significant amount of positive nutrient leverage include:
Remember how I said nutrient density is the icing on the cake (but protein is the cake)? This analysis demonstrates that:
- by itself, increasing protein provides a 17% increase in satiety, however
- together, dialling in nutrient density will give you a further 11% benefit in satiety.
Swapping more fibrous carbohydrates for refined carbohydrates is beneficial (1.8% reduction in calories), but prioritising protein (17%) and nutrient density (11%) is much more powerful.
So is it “protein leverage” or “nutrient leverage”?
- Overall, the data indicates a statistically significant nutrient leverage effect at play. Foods that contain more protein also tend to contain more of the other essential nutrients.
- Some of the satiety impacts generally attributed to protein can also be attributed to our cravings for other essential nutrients.
- The satiety response to a particular nutrient is dependent on the current intake relative to the optimal requirement for that nutrient.
Good data analysis can give you powerful insights that enable you to achieve your goal with a precise prescription. But we need to keep in mind that we are still limited by the availability of comprehensive data for our food. While we have solid data for macros (protein, carbs and fat), the data is still sketchy for individual micronutrients.
For example, we know that increasing protein % leads to a significant decrease in calories consumed. We also know that protein contains nine essential amino acids that all independently align with greater satiety (as shown in the chart below).
However, I can’t show you a multivariate analysis showing the satiety benefit of each amino. Even with 125,761 days of data from 32,192people, we don’t have enough good data on amino acids because they are not regularly measured.
Similarly, I hypothesise that if we had accurate data for all of the micronutrients, I would be able to show you a statistically beneficial effect for each of the vitamins and minerals. But because we don’t have comprehensive data for every food logged by Optimisers, we only see a significant relationship for the micronutrients most commonly measured in food.
We must be even more careful when using other parameters (e.g. phytochemical and flavonoid content) where data is limited or subjective parameters can’t be easily validated and calibrated.
While all this data analysis is fun, we need to be mindful of the limitations when designing a food ranking system. Adding excessive parameters can lead to data curve fitting and some nonsensical outliers.
Context is important. Protein is unlikely to be as crucial for carnivores or people following a low carb diet. They might need to focus more on getting some of the vitamins and minerals. Meanwhile, people on a vegan or vegetarian diet may be struggling to get adequate B12, omega 3 or bioavailable protein.
There is no one size fits all prescription for everyone.
When applying nutrient density at the individual level, we want to use the data to help people get more of the nutrients they are currently not getting enough of to create a well-balanced diet at a micronutrient level, regardless of where those nutrients come from. That’s why we created the micronutrient fingerprint to gamify the process of Nutritional Optimisation.
The chart below shows the micronutrient fingerprint for Karen’s 100% ONI score in our Nutritional Optimisation Masterclass. Karen happens to be following a vegetarian diet and is VERY motivated to optimise her diet, but in general, we usually see people get the best results on an omnivorous diet. For more details on what Karen eats and her amazing story, check out this article.
The black line represents 100% Optimal Nutrient Intake for each of the essential micronutrients. If you can fill the area to the left of the black line (without fortified foods or supplements), you get a perfect 100% score. To do this, you have to keep chasing foods that contain more of the nutrients you are currently lacking.
In our Masterclass, Nutrient Optimiser gives precise food and meal recommendations that help people dial-in both nutrient density and satiety.
Again and again, beyond protein leverage, we see that people who pursue a higher ONI are satisfied with fewer calories. The chart below shows the ONI score vs calorie intake, showing that, as people increase the nutrient density of their diet, they tend to eat less.
If you’re interested in a challenge, the table below shows the Optimal Nutrient Intakes (ONI) that align with the lowest calorie intake or the 85th percentile intake from the Optimiser data.
For comparison, I’ve also shown the Dietary Reference Intake (DRI) levels which are the minimum intake levels required to prevent disease related to malnutrition. Ironically, they also align with the lowest satiety outcome based on the nutrient-satiety analysis. Sadly, many people even struggle to meet the DRI in our modern food system.
|Vitamin E||2333||10000||IU/2000 cals|
|Vitamin E||15||34||mg/2000 cal|
|Vitamin D||600||3700||IU/2000 cals|
|Vitamin D||75||400||mg/2000 cal|
|Thiamin (B1)||1.1||6.5||mg/2000 cal|
|Riboflavin (B2)||1.1||6||mg/2000 cal|
|Niacin (B3)||14||70||mg/2000 cal|
|Pantothenic acid (B5)||5||12||mg/2000 cal|
|Vitamin B6||1.3||5||mg/2000 cal|
|Vitamin B12||2.4||35||mg/2000 cal|
|Vitamin K1||90||1100||mg/2000 cal|
|Omega 3||1.1||6||g/2000 cals|
Comparison of ONI score vs Food Compass
Finally, I’ve plotted the Food Compass Score vs the Optimal Nutrient Intake Score for some popular food to see how the two ranking systems compare. For more detail, you can check out the interactive Tableau version here.
Fats & oils
It’s interesting to note how olive oil and peanuts are practically ranked as superfoods by Food Compass (due to the plant oil and nuts and seeds parameters) but don’t do nearly as well in the ONI score.
The chart below shows animal and seafood based foods. In the top right corner we see that both systems rank seafood highly. However, the big difference is with the animal products. Due to a lack of emphasis on protein and the use of saturated fat and cholesterol as negative ranking factors, Food Compass ranks meat and dairy significantly lower.
When we look at plant-based foods, both systems rank non-starchy vegetables highly (shown in the top right). We see that Food Compass ranks fruits and nuts much more generously towards the lower right, despite a lower nutrient density.
If you want to dive into this data in more detail, you can check out the interactive Tabluea version here.
If you’ve made it this far, chances are you’re pretty serious about your nutrition.
If you’d like to see how your current diet stacks up, we’d love you to take our Free Food Discovery Challenge here. After a few days of tracking, you’ll see how your diet compares with other optimisers and discover what foods and meals you need to fill your micronutrient gaps.
We’re on a mission to gamify nutrition. We love it when we see people lose themselves in playing the game. Before long, they’ve not only climbed the leaderboard, but they’ve also lost a ton of weight (without really trying) and feel great, energised by the nutrients their body needs to thrive!
So, do you think the Food Compass System is broken?
Which system do you think provides more logical results and why?
I’d love to hear your thoughts in the comments below.
How Can I Calculate My Nutrient Intake?
Level Up Your Nutrient Density
To help you level up your nutrient density, we’ve prepared a Nutritional Optimisation Starter Pack to ensure you are getting plenty of all the essential nutrients from the food you eat every day.
The free starter pack includes:
- Maximum Nutrient Density Food List
- Sample Maximum Nutrient Density Recipe Book
- Sample Maximum Nutrient Density Meal Plan.
To get started today, all you have to do is join our new Optimising Nutrition Group here.
Once you join, you will find the Nutritional Optimisation starter pack in the discovery section here.
- What is nutrient density?
- The biggest trends in nutrition
- Protein for weight loss: How much you need and why it works
- Does eating fat make you fat? The surprising truth about cholesterol and saturated fat!
- We want to be the Google for nutrition!
- How to eat to save the planet (and yourself)
- What Do Top Optimisers Eat Across the Globe?
- 7 Day Nutrient Clarity Challenge