“Weight gain is about eating too many calories, and weight loss is about finding a way to eat fewer without being miserable.” Herman Pontzer
Over the past five years or so, I’ve been working to develop the ultimate satiety algorithm.
Why?
Once we understand the factors that align with eating less, we can reverse engineer food and meal choices to empower Optimisers to imitate people who successfully maintain a lower energy intake and the many benefits that accompany it.
But…
As I’ve dug into the data, I’ve realised that no there is no single satiety algorithm optimal for every dietary approach and dietary preference.
The solution…
For even better results, we can use the satiety ranking factors tailored for specific dietary approaches.
In this article, I want to show you how you can optimise your food and meal choices for greater satiety, regardless of whether you prefer:
- low carb,
- low fat,
- high protein, or
- low protein.
Executive Summary
Our data analysis shows that, while protein and energy density dominate the satiety equation, we can refine the satiety algorithm further by emphasising micronutrients that tend to be harder to find with specific dietary approaches.
Building on this understanding, we can reverse engineer our food and meal choices to achieve greater satiety.
By emphasising foods and meals that contain more of the priority nutrients, we can identify foods and meals that contain more of all the essential nutrients we need to thrive and feel satisfied.
If you just want to know what to eat without diving into the analysis, you can access our free food lists, meal plans and NutriBooster recipes in our Optimising Community by clicking the links below.
| Approach | Foods | Recipes | Meal plan |
| General | Access | Access | Access |
| Low carb | Access | Access | Access |
| Low fat | Access | Access | Access |
| High protein | Access | Access | Access |
| Low protein | Access | Access | Access |
| Low satiety | Access | NA | NA |
| Nutrient Dense | Access | Access | Access |
To learn more about the analysis behind these unique lists, read on.
Satiety Per Calorie vs Nutrient Density
Satiety and nutrient density are related but different.
- Nutrient-dense foods and meals provide more of all the nutrients per calorie. We quantify nutrient density using our diet quality score.
- Satiety is the absence of hunger and the desire to eat. To lose weight and improve metabolic health, we need to prioritise foods and meals that provide greater satiety per calorie.
Together, satiety and nutrient density empower us to feel satisfied for longer without excess energy while also getting all the amino acids, minerals and vitamins we need to feel energised and thrive.
The Data
This analysis is based on 304,311 days of data from free-living humans (just like you).
A little over half comes from the NHANES nutritional surveys, while the rest comes from people logging their food in Cronometer and using Nutrient Optimiser.
As you will see, this data, from thousands of people following a broad range of dietary approaches, helps us understand the nutritional factors that align with eating less vs eating more.
This is not a metabolic ward study where every bit of food is perfectly controlled. But we have 833 person-years of data that enable us to identify the statistically significant parameters that explain how much we eat.
I hope that one day someone will throw a lot of funding to validate this approach to quantifying satiety in a randomised control trial.
But if you can’t wait, this data analysis provides highly actionable insights to optimise your food and meal choices to align with your goals and feel satisfied with less energy.
As you will see from the food lists and images below, the outcome seems to make sense.
Multivariable Analysis
Most people eat when hungry and stop when they’re full, eating to satiety.
With data on how people eat in the real world, we can use multivariable regression analysis to identify the key nutritional parameters that align with higher vs lower energy intake.
With a large amount of data, we achieve a very high degree of statistical significance (i.e. these relationships are not due to chance). The highest p-value in this analysis is 0.0000000000005 (or 5E-13), which you’d never see in a clinical study.
The regression coefficients from the analysis allow us to estimate how much moving from low to high amounts of these parameters aligns with a lower energy intake.
When we bring this together, we can estimate how much we would eat of a certain food or meal based on its nutritional properties.
In previous articles, I’ve looked at the 304,311 days of data as a whole as well as breaking it down into the following subsets, with 100k days of data in each:
- Low carb,
- Low fat,
- High protein, and
- Low protein.
For further background details, see:
In this article, I want to bring it all together to highlight the unique satiety factors for each approach.
General Approach
First, let’s look at the results when we look at all the data together. The table below shows the statistically significant satiety factors when considering all 304,311 days of data.
| Nutrient | Calories |
| protein (%) | -476 |
| energy density | -173 |
| calcium | -101 |
| potassium | -89 |
| Iron | -70 |
| fibre (%) | -63 |
| vitamin C | -19 |
The ‘calories’ column is the change in energy intake that occurs when we move from one extreme to another when the other statistically significant factors are considered. For example, moving from 12% protein to 38% protein (i.e. 15th percentile to 85th percentile for protein) aligns with a 476-calorie reduction.
In previous articles, I’ve shown the results of the multivariable analysis, which some people have told me they find confusing. So, to simplify, the figure below shows the reduction we can attribute to each satiety factor based on the analysis. When we consider all the data, we see that protein % is the dominant factor explaining the variance in how much we eat, but there are other factors.
Alignment With Other Research
The fact that protein is the dominant factor aligns with the protein leverage hypothesis, first proposed by Professors David Raubenheimer and Stephen Simpson in 2005.
But the fact that protein only explains 50% of the variance aligns with findings by Kevin Hall from his more recent analysis of metabolic ward studies.
The fact that we see a large portion of calorie intake explained by energy density aligns with research by Professor Barbara Rolls.
Intriguingly, Professors Raubenheimer and Simpson have noted that we have appetites for other nutrients beyond protein, like calcium and sodium, but there are likely others.
Nutrient Leverage?
When we consider all three hundred thousand days of data, the analysis indicates that we may have an appetite for other nutrients, including calcium, potassium, iron, fibre, vitamin C, folate, sodium and selenium.
While these micronutrients have less impact than protein or energy density, they help explain more of the variation in energy intake and improve the accuracy of our satiety scoring system in addition to using protein or energy density alone.
Research on the satiety response to specific nutrients is still sparse, but there is some.
- Sodium is a critical mineral that we have a strong conscious craving for. We crave salty foods to get enough sodium. But when we’ve had enough, our food starts to taste too salty (Shulkin, 1991).
- Calcium is critical to maintaining our bones. In Calcium: taste, intake, and appetite (2001), Tordoff showed changing taste perceptions for calcium to maintain homeostasis in the blood and body. Calcium, along with potassium (which is a satiety factor), is considered by the US Dietary Guidelines as a nutrient of concern that most Americans aren’t getting enough of.
- Vitamin C was the first micronutrient discovered by James Lind to cure scurvy in the first randomised controlled trial (RCT) in the 1860s. When sailors got hold of the fruit they had been craving, they swallowed it ‘with emotions of the most voluptuous luxury’.
The fact that minerals like potassium and calcium tend to feature as dominant satiety factors may be because they are larger is because they are less frequently used in supplements and food fortification. These nutrients have also declined in our food system since the introduction of chemical fertilisers used in the 1930s to turbocharge industrial agriculture.
Meanwhile, the fact that vitamins are added to ultra-processed foods means that the data for nutrients we only need in smaller amounts is noisier and thus not statistically significant in the multivariable analysis. Thus, the analysis doesn’t prioritise ultra-processed foods that are fortified to compensate for their nutritionally bankrupt ingredients (i.e. refined grains, sugar and industrial seed oils).
Rather than nutrient leverage, another way to view this is that we are identifying the signature of high-satiety foods and meals. Similar data-driven techniques are the core of machine learning used by Artificial Intelligence algorithms to identify cancer with CT scans better than radiologists.
Sadly, bias, preferences and beliefs cloud human judgement in nutrition. A data-driven approach is agnostic to food preferences and identifies the signature of foods that tend to provide greater satiety per calorie for most people most of the time.
High Satiety Foods and Meals
Rather than just being a ‘fun fact,’ we can use the regression coefficients from the analysis to calculate how much we are likely to eat of a specific food and thus assign a satiety score.
The image below shows the satiety scores of some popular foods. You can find a longer list and printable .pdfs in our Optimising Community here.
We can also apply the same approach to meals and recipes. The image below shows some examples of high-satiety NutriBooster recipes.
The micronutrient fingerprint chart shows the nutrient profile of the top 300 (of 1400) NutriBooster recipes when we prioritise these statistically significant nutrients. The priority nutrients are shown in red.
Notice that we not only get more of the priority amino acids and micronutrients, but we also get a solid amount of all the other nutrients that come packaged with these key nutrients.
Low-Satiety Foods
We can also use the satiety algorithm to identify foods that provide the lowest satiety. The figure below shows some examples. For more details, check out the full lists here.
We all need some energy in our diet, so there’s nothing wrong with these foods in moderation. However, if you want to lose some excess weight, you might benefit from prioritising the higher-satiety foods while dialling back these lower-satiety foods.
High-Satiety, Lower-Protein Foods
When I analysed the NHANES data, I found that many Americans consume a lot less protein than our Optimisers, with an average protein intake for the NHANES data. Nearly 10% of Americans do not meet the 10% minimum recommended by the US Dietary Guidelines.
The chart below shows the satiety response to protein % using all the data.
- We eat and store the most when our diet contains about 12% protein.
- To the right, we can see that for the ‘high protein’ subset (i.e., more than 26% protein), the satiety response to a higher protein percentage is essentially linear—a higher protein percentage aligns with eating a lot less.
- However, for the ‘low protein’ subset to the left (less than 15.4% protein), increasing protein % aligns with a higher energy intake.
Thus, the protein leverage model breaks with low protein intakes. This explains why some people find a very low-protein diet satiating (e.g., fruitarian, therapeutic keto, whole-food plant-based diets, and traditional subsistence diets of rice or tubers).
The chart below shows the statistically significant satiety factors for a lower protein diet.
We’ve mentioned calcium, potassium and vitamin C above, but it’s worth noting here that, according to the World Health Organisation, iron is the most common nutrient deficiency worldwide, particularly in poorer counties, due to limited access to higher-protein animal-based foods. Iron is also much less bioavailable from plant-based foods than the heme iron in higher protein animal-based foods.
Again, using the regression coefficients for the low-protein data, we can define a unique satiety score for low-protein foods. You can download longer printable food lists in our Optimising Nutrition Community here.
We can also identify high-satiety, low-protein NutriBooster recipes that look like this.
But before you jump to a low-protein diet, it’s worth pointing out that the low-protein subset of the data has the highest energy intake and the lowest nutrient density of all the extremes. Avoiding protein will likely make it harder to get several nutrients. For more details, see Micronutrients at Macronutrient Extremes.
The micronutrient fingerprint below shows the nutrient profile of the top 300 recipes when we rank using the low-protein satiety factors. While we still get a solid amount of most of the micronutrients, it’s not as nutrient-dense as the other approaches discussed in this article.
There’s nothing wrong with these foods and meals if you’ve already met your daily protein and micronutrient quota. These foods and meals have a lower energy density and heaps of fibre, so they’ll still be hard to overeat. If you start the day with a solid dose of protein and nutrients, you can afford to backfill with some energy later in the day, especially if you’re not trying to lose weight aggressively.
High Protein
Unsurprisingly, protein is the dominant satiety factor for the high protein data. However, fibre, calcium, potassium, folate, sodium, energy density, and selenium also play a role in the satiety equation.
Foods that contain more of these other nutrients per calorie tend to be non-starchy vegetables that complement the nutrient profile of meat and seafood. They also boost satiety by reducing energy density and adding some bulk.
The image below shows a selection of popular higher-protein foods and their satiety scores using the high-protein ranking. You can download longer printable food lists in our Optimising Nutrition Community here.
Food lists are great. But we don’t just eat single foods. We combine them into meals. Below are some examples of our NutriBooster recipes that rank well using the high satiety, high protein satiety equation.
The micronutrient fingerprint below shows the nutrient profile of the top 300 recipes (of 1400) when ranked by using high-protein data.
Lower-Carb, High-Satiety Foods
Similar to protein, the data shows that the satiety response to carbohydrates is not linear.
We tend to eat the most when our food contains around 40% non-fibre carbohydrates, with most of the remaining energy coming from fat. This combination of fat+carbs is the signature of ultra-processed foods we can’t stop eating.
Towards the left, we can see that reducing non-fibre carbohydrates from 40% to 14% corresponds with a 30% reduction in energy intake. But towards the right, a low-fat diet with most of the energy from carbohydrates is also hard to overeat.
So we can’t simplistically say carbs/fat = good/bad. It’s a bit more nuanced than that.
The chart below shows the satiety factors for the low-carb data (i.e. less than 21% non-fibre carbohydrate). Again, protein is the dominant satiety factor. But people on a lower-carb diet will eat even less if they get more fibre, folate, potassium, sodium and calcium per calorie from their food.
The image below shows high-satiety foods using the low-carb satiety score. If you prefer a lower-carb diet, you can download the list of higher-satiety, lower-carb foods in our Optimising Nutrition community here.
The image below shows some examples of our NutriBooster recipes that rank using the low-carb satiety criteria.
The micronutrient fingerprint chart below shows the priority nutrients that are harder to find on a lower-carb diet (e.g. calcium and folate). Note how calcium and folate are high-priority nutrients and harder to get on a lower carbohydrate diet unless you also emphasise non-starchy green vegetables.
Lower-fat, High-Satiety Foods
Finally, we come to the analysis of the low-fat data (i.e. less than 32% fat).
Protein is still the dominant satiety factor, and people are eating a lower-fat diet. However, energy density has a much bigger role because protein tends to be lower. Calcium, fibre, potassium, iron and vitamin C help us further refine the satiety equation to explain the variance in energy intake on a lower-fat diet.
Again, we can use these satiety factors to identify high-satiety food for people who prefer a lower-fat diet, and you can view and download the full list of popular food here.
We can also use the satiety factors to rank our NutriBooster recipes, as shown in the examples below.
The micronutrient fingerprint for the top 300 recipes ranked with the lower fat satiety equation show that we still get a solid amount of nutrients. However, if you compare the fingerprint charts, you’ll see that some nutrients are easier to get on a lower-fat diet vs low-carb (and vice versa).
Which Approach is Best?
By now, you may be wondering, ‘Which approach is best?’
I tried to answer this by looking at the top 300 NutriBooster recipes with the lowest forecast calorie intake using each approach. I’ve also added the 300 recipes with the highest diet quality score. The results are shown in the chart below.
- The refined satiety factors for low carb, low fat, and high protein give us a slightly lower forecast energy intake than simply using the ‘one size fits all’ approach.
- Even though they have different nutrient profiles, the low-carb and low-fat approaches provide a similar diet quality score.
- The high-protein recipes have the lowest energy intake. However, you may not need to wind up your protein % that much to achieve sustainable long-term results.
- The low protein approach has the lowest diet quality score and a higher energy intake.
- The maximum nutrient density recipes provide the best diet quality score and a slightly higher forecast calorie intake.
Ultimately, the best diet is the one you will enjoy and stick to. So, for aggressive weight loss, you could choose the high-satiety low-fat, low-carb, or high-protein approaches. Meanwhile, the maximum nutrient density approach will also be highly satiating while packing in even more nutrients per calorie!
Further Personalisation
In this article, we’ve demonstrated that rather than creating a one-size-fits-all satiety model, food and recipes, we can tailor the satiety algorithm to different dietary approaches. We may crave some nutrients that can be harder to find with different dietary approaches.
All these approaches build on a solid foundation of protein. Hence, ensuring you have adequate protein without excess energy from carbs and fat is a great place to start your journey of Nutritional Optimisation to increase satiety. This is the process we guide Optimisers through in our Macros Masterclass.
While we can divide the data into subsets to identify the priority nutrients, identifying your priority nutrients is even more powerful. If you wanted to take this a step further, you could review your current diet to identify your priority nutrients and add in foods and meals that contain more of them. This is the process we guide Optimisers through in our Micros Masterclass.
If you want to identify your priority nutrients, you can take our Free Nutrient Clarity Challenge.
Summary
- Protein and energy density are the dominant factors that align with greater satiety. However, micronutrients provide greater resolution to increase satiety further.
- The weightings of the various satiety factors vary for different dietary approaches. We can use these to optimise satiety for food and meal choices further.
- But if you want to dial things in even more, you can track your diet and identify the foods and meals that will fill the gaps. Simple food lists are the best place to start for most people.
- However, for most people who usually don’t want to track their food, the simple list of foods and nutrient-dense recipes is the best place to start their journey of Nutritional Optimisation.
| Approach | Food list | Recipes | Meal plan |
| General | Access | Access | Access |
| Low carb | Access | Access | Access |
| Low fat | Access | Access | Access |
| High protein | Access | Access | Access |
| Low protein | Access | Access | Access |
| Low satiety | Access | NA | NA |
| Nutrient Dense | Access | Access | Access |