Calculating Satiety Per Calorie: A Data-Driven Approach to Nutrition

Different dietary approaches, whether low-carb, high-protein, carnivore or plant-based, become popular because they help people:

• increase satiety
• feel full and satisfied
• consume fewer calories, and
• achieve weight loss.

Unfortunately, dietary approaches based on the avoidance of ‘bad things’ in food (e.g., fat, carbohydrates, animal-based foods, plant-based foods, sugar, salt, cholesterol, polyunsaturated fat, saturated fat, etc.) create a limited selection of foods that many find hard to sustain. 

Moreover, these approaches can be manipulated to make foods more palatable while still following the rules, leading followers to stall out and often regain their initial lost weight (e.g., paleo comfort foods, keto fat bombs, and ultra-processed plant-based foods).

Satiety per calorie is a novel approach that uses quantifiable nutritional factors to help users improve their diets for lasting satiety.

In this article, we explore the development and latest version of our data-driven satiety algorithm.  You’ll learn:

• Why certain foods make us feel ‘addicted’ and lead to overeating
• How other foods can satisfy our cravings with fewer calories.

This article is long and detailed, so you can use the table of contents below to jump to the sections you’re most interested in.  But if you simply want to understand what to eat to increase satiety, check out the most satiating foods article here.

Definitions

Before diving into the details of our satiety algorithm, let’s define a few key terms to ensure a common understanding. 

Satiety

Satiety is the feeling of fullness and satisfaction that occurs after eating, reducing the urge to eat for a period.  Foods with high satiety can help control hunger and maintain a healthy diet by providing the necessary nutrients without excessive calories.

Satiety can be achieved with any food, regardless of the amount of energy consumed.  Unfortunately, when we’re ravenous, our natural response is to satiate our hunger quickly with energy-dense foods like pizza, chocolate or ice cream, which often provide more energy than our bodies require to be metabolically healthy.

Satiation

Satiation is the point in time when we feel full and satisfied enough to stop eating. 

A large, low-energy-density meal, like a watery soup or large salad, may help us feel full and satiated with minimal energy.  However, it may not keep us feeling full for long as a fatty steak with plenty of protein and energy.

Subjective Satiety

Subjective satiety is how satiated we feel after a particular meal. 

For example, someone might say, “I feel satiated with a big fatty steak, but I feel hungry a few hours after eating all the skinless chicken breast I can stomach.” Here, they are making a subjective judgement of their sensations of satiety with no concept of how much energy was consumed.

Unfortunately, again, subjective satiety does not consider how much energy was consumed to achieve satiety.  It’s likely that if all we had to eat was skinless chicken breast, we would consume a lot less energy across the day than from the fatty steak, even if we ate more often. 

Sensory Specific Satiety

Sensory-specific satiety occurs when we get as much as we need of a particular nutrient. 

As we eat more of a meal, we satisfy our cravings for the nutrients it contains; thus, the pleasantness decreases. 

Although we can consume all the lean steak or chicken breast we can stomach, we always have room for dessert.  Dessert balances the lean protein with energy from fat and carbs. 

Our bodies are constantly working to find the balance of nutrients and energy in our food environment with the least energy expenditure (known as Optimal Foraging Theory).   

• We crave salt, but if we add too much salt to a meal, it will taste ‘too salty’, and we’ll eat less of it. 
• We have an intense craving for sweet-tasting foods, but if we add too much sugar to our food, it can taste ‘sickly sweet.’ 
• We crave protein, but our bodies have limited capacity to store amino acids, so once we get the protein we need, our appetite for protein-rich foods shuts down. 

Specific Appetite

Specific appetite refers to the craving for foods that contain particular nutrients to meet the body’s physiological needs. 

Humans have an unlearned appetite and conscious taste for sodium (salty taste), sugar (sweet) and protein (umami).  However, we can also subconsciously develop a ‘learned appetite’ through association with the effects of certain foods and their beneficial effects over time. 

Some research on animals and humans has focussed on our specific appetite or nutritional wisdom for other micronutrients.  Cravings and specific appetites are more challenging to study due to the need to induce nutrient deficiencies at scale.  However, as detailed in this article, our satiety analysis suggests that we crave all the essential nutrients to varying degrees and will eat more food to meet our minimum requirements. 

Bliss Point

While we often think linearly (e.g. good vs bad, black vs. white, high vs. low), many things in nature follow a non-intuitive inverted U-shaped curve, including our appetite and cravings.    

The bliss point is the concentration of a particular nutrient where our craving changes to satiety.

Howard Moskowitz pioneered the quantification of the bliss point in food in the 1970s, paving the way for the processed food industry to optimise products with the perfect blend of sugar, fat, and salt to maximise consumption and profit.

As shown in the centre of the satiety response curves below, the bliss point aligns with the concentration of each nutrient where we eat the most.  It’s the perfect blend of that nutrient with energy that maximises palatability. 

• When a food contains very little of the nutrients we need, it tastes bland, and we eat less. 
• If a food has a lower concentration of the nutrients we need, we keep eating until we get enough.
• When our food contains a higher concentration of all essential nutrients, we satisfy our cravings with less energy.

A meal or food that hits multiple bliss points produces a supraphysiological dopamine response, so we’ll eat more of it.  Humans have always used technology to combine the nutrients they need to survive and store enough fat to survive the coming winter — from hunting to cooking with fire to Grandma’s secret Thanksgiving recipe — but modern food technology has perfected this with engineering precision, giving us these foods 365/24/7. 

Optimal Nutrient Intake

The Optimal Nutrient Intake (ONI) is our stretch target concentration for each nutrient that triggers sensory-specific satiety more efficiently, with less energy. 

We have set the ONIs at the minimum of:

• three times the bliss point, or
• the point at which the satiety response tapers off, and more of that nutrient provides negligible additional benefit.   

While the bliss point generally aligns with the Dietary Reference Intakes, targeting the ONIs enables us to get the nutrients we need with less energy.  Working towards the ONIs for all the essential nutrients, as we do in the Micros Masterclass, enables us to trigger sensory-specific satiety with multiple nutrients.

The ONIs also define the point of diminishing returns for the essential nutrients, beyond which we should focus on other priority nutrients.  

Additionally, in Sensory influences on food intake control: moving beyond palatability, Forde and McCrickerd suggest that nutrient-dense foods tend to have a more robust, fuller taste, signalling to our bodies that we don’t need as much of them to get the nutrients we need from them.

Satiety Per Calorie

Satiety per calorie (SPC) is a measure of the satiety provided by foods or meals relative to their caloric content.

SPC helps individuals understand which foods will help them feel full and satisfied while consuming fewer calories throughout the day.  Thus, it can be beneficial for weight management and overall metabolic health.

As detailed below, we have quantified satiety per calorie using 655,283 days of food intake data from people following a wide range of dietary approaches all over the world.

Satiety Score

Our satiety algorithm allows us to estimate how many calories of a particular food or meal someone would consume per day if that’s all they had to eat.  We can then rank food and meals on a scale from 0 to 100% (i.e. the satiety score).  We will consume more food with a satiety score of 0 and less of those with a satiety score of 100%. 

While the highest-satiety foods may seem self-evidently healthy to many, quantifying SPC helps us identify better or worse options from all food available based on individual goals, context and preferences without resorting to named diets. 

SPC allows us to move beyond named diets (e.g. keto, plant-based, low-fat, etc.) to compare foods based on their nutritional properties. 

Previous Attempts to Quantify Satiety

The latest iteration of our satiety algorithm builds on several previous attempts to quantify satiety, which are briefly summarised below. 

Satiety Index of Common Foods

The University of Sydney study, A Satiety Index of Common Foods (Holt et al., 1995), quantified the satiety response to 1000 kJ portions of 38 common foods.  They measured subjects’ subjective satiety over two hours and calculated the area under the curve changes in satiety.   

As we can see in the chart from the study below, cooked and cooled plain potatoes scored the highest, and croissants scored the lowest.  In general, processed foods (tagged as hyperpalatable refined fat + carbs at the bottom) scored the worst. 

Despite being nearly thirty years old, this is still the most referenced study in the satiety space.  Unfortunately, no one has gone back since to expand this dataset in a controlled study.  Hence, it’s challenging to extrapolate this small dataset to a broader range of foods, especially as only a couple of high-protein foods were tested.

The other major issue with this study is that it only tested the satiety response over two hours. In a companion paper (Interrelationships among postprandial satiety, glucose and insulin responses and changes in subsequent food intake, Holt et al. 1996), the authors noted that foods that spiked insulin and glucose tended to provide more short-term satiety. 

But we can’t extrapolate this 2-hour test to understand how much we might eat in a day or a week.  For many, the reactive hypoglycaemia caused by refined carbohydrates might make them hungry two or three hours later. 

Energy Density

Energy density is another satiety heuristic popularised by Professor Barbara Rolls, who started studying the relationship between energy density and energy intake early this century.  The concept has gained popularity in the low-fat plant-based communities.  Unfortunately, similar to the satiety index study, Rolls’ studies only measured satiety over two hours.   

As you can see in the chart below, from our satiety analysis, low-fat, low-energy-density fruits and vegetables do align with eating less compared with processed foods. But towards the left, we see that high-protein and high-fat foods don’t follow the linear trend. 

Our data also shows that few people eat a diet that is low enough for energy density to be a dominant satiety factor over the long term.  Most people on a plant-based diet add dressings and oils to their vegetables, resulting in an energy-dense, low-satiety diet. 

Protein Leverage

Protein leverage is a critical factor in diets, particularly low-carb diets.

This concept, proposed by University of Sydney professors David Raubenheimer and Stephen Simpson in 2005, suggests that protein intake significantly influences satiety.  Our analysis also shows that increasing the percentage of protein in your diet can help reduce overall calorie consumption and enhance satiety.

While initially understood as a linear relationship (as shown in the chart above), Raubenheimer and Simpson’s recent work shows that there is a breakpoint in the protein-satiety response, with energy intake dropping once we fall into the protein malnutrition zone.   

Our data analysis also shows that we eat less if our food contains minimal protein but eat the most when our food contains around 12.5% protein, which appears to be the optimal protein concentration required to maximise growth and fat gain and minimise dietary-induced thermogenesis. 

As we move to the right of the chart, we get a higher concentration of protein per calorie (i.e. a higher protein percentage) and eat less because we satisfy our cravings for protein with less energy. 

When we consider our entire dataset, moving from 12.5% to 50% protein aligns with a 34% reduction in energy, but this increases to a 45% reduction in energy intake if we only consider the lower carb data.   

In our Macros Masterclass, we guide our Optimisers to incrementally increase their protein % to increase satiety and achieve sustainable weight loss.  We use our satiety algorithm to show them how they could tweak their current food intake to increase satiety with a limit of +10% on their protein percentage.  Large jumps in protein % tend to lead to increased hunger due to the massive drop in energy intake, making the weight loss process unsustainable. 

If we draw a simple linear regression through the data, we get an R2 correlation coefficient of -21.9% (p < 0.000000005).  So, there is no doubt that protein leverage is a crucial factor in the satiety equation.  Still, it explains less than a quarter of the variation in our energy intake. 

Raubenheimer and Simpson have noted that humans possess a specific appetite for at least five nutrients: protein, carbohydrates, fat and at least two micronutrients: salt and calcium.  However, in their 2022 paper, An Integrative Approach to Dietary Balance Across the Life Course, they noted that specific appetites for other nutrients likely exist. 

Due to a lack of data, specific appetite for additional nutrients has yet to be identified or rigorously quantified.  Unfortunately, it is not feasible or ethical to undertake controlled studies where we intentionally induce nutrient deficiencies to test our cravings. 

Multiple Linear Regression

In an effort to develop a more accurate satiety system for our Optimisers, I started working on a quantified approach to modelling satiety back in 2018.  Initially, I used 587,187 days of MyFitnessPal data, which unfortunately only contained macronutrient data.

Once we had collected enough data from people using Nutrient Optimiser, I started investigating the impact of micronutrients, hoping to identify other nutrients for which we have a statistically significant specific appetite.  

When I chatted with David Raubenheimer and Simpson on my podcast in 2021, they suggested I consider a multivariate regression analysis, or multiple linear regression.  Multivariate regression analysis allows us to identify the statistically significant factors that align with eating less vs eating more, eliminate the non-significant factors and determine specific weightings for each of the remaining factors. 

The bar chart below shows the satiety factor weightings determined using the linear regression approach, with an R2 correlation of 28.8% achieved with measured energy intake.

This approach, which is a significant improvement on protein leverage alone, allows us to predict 28.8% of the variation in energy intake without knowing anything about a person’s weight, gender, activity levels, or muscle mass. 

Since then, I’ve kept in touch with professors Raubenheimer and Simpsons and periodically shared my data and analysis with them.  They have been generous with their time and guidance.  In late 2023, I added the NHANES dataset (159,267 days) to our data from Optimisers. 

In March 2024, I talked with Dr Alistair Senior, a colleague of Raubenheimer and Simpson.  At their request, he re-ran my analysis and came to the same conclusion.  That is, rather than merely protein leverage, there also appears to be a nutrient leverage effect.

Expanding the Nutrition Dataset

In our chat, Dr Senior also mentioned that he had run some quadratic regressions to model the inverted U-shaped craving/satiety responses to each nutrient.  He also mentioned an FAO/WHO dataset, which contains data from lower and middle-income counts that we could add to the dataset to refine the analysis before working towards the publication of our findings.

So, I downloaded the FAO/WHO data and added it to our dataset.  The table below shows the sources of the food intake data that we now have, spanning a wide range of eating patterns from different cultures all over the world.  

Source	calories	protein (g)	carbs (g)	fat (g)	fibre (g)	days
Optimisers	1474	111	84	77	20	191,632
NHANES	1877	71	237	71	14	159,267
Brazil	1688	75	235	50	19	141,994
India	1622	45	290	31	30	39,920
Philippines	1404	49	223	35	3	29,910
Tunisia	2104	69	321	61	31	28,580
Mexico	1787	57	253	61	14	10,185
Italy	2113	82	262	82	18	9,465
Kenya	1643	45	294	32	26	6,543
Romania	2094	102	261	75	31	6,539
Argentina	1695	69	213	63	11	6,441
Bangladesh	1753	47	315	34	15	4,381
Bulgaria	1293	43	163	52	13	4,040
Guatemala	2152	66	370	45	31	2,933
Burkina Faso	1742	79	308	26		2,605
Uganda	1885	52	343	34	22	2,217
Malaysia	1774	63	242	62	8	1,629
Nigeria	1786	59	255	59	17	1,349
Tanzania	1200	31	196	32	27	1,297
Mozambique	1381	35	228	37	22	816
Peru	986	34	161	23	0	619
Costa Rica	1956	64	301	55	14	566
Pakistan	1451	37	243	37	19	515
Cabo Verde	1615	73	178	68	16	484
Indonesia	2154	72	338	57	0	360
Bolivia	1772	61	298	38	18	328
Ethiopia	1708	36	262	57	17	323
Congo	1864	40	256	76	22	271
Cameroon	1555	39	219	58	0	74
All	1670	80	112	61	18	654,476

Twenty-eight per cent of this data is from our Optimisers, who record their data in Cronometer, which is then imported via API into Nutrient Optimiser.  While this data is ‘self-reported,’ it is helpful to understand what happens at the higher nutrient density, lower carb, and higher protein end of the spectrum. 

The NHANES dataset is managed by the US CDC, while the FAO/WHO dataset is gathered from controlled surveys and studies undertaken by universities and governments.  The broad range of data sources enables us to understand the satiety response to a more comprehensive range of macronutrients and micronutrient intakes.  For more details on the data, see:

• How We’re Using Big Data to Help Anyone Eat Better and
• Global Nutrition Trends: Protein, Carbs & Fat Around the World.

While no measured data will be perfect, the size of this combined dataset gives us some exciting opportunities to better understand our specific appetite and sensory-specific satiety response to individual essential nutrients.

Our Latest and Greatest Satiety Scoring System!

This section discusses the most recent refinements made to our satiety model with the expanded dataset. 

Two Way Linear Regression

I have previously charted the satiety response to each of the essential nutrients using the GamFIT function in RStudio.  Appendix B contains the charts for the shortlist of nutrients incorporated into the model.   

While these charts provide an illuminating graphical representation of the relationship between the concentration of each essential nutrient and how much we eat, the GamFIT function does not give a formula that can be applied to other foods.

After talking to Dr Senior, I initially tried to model the satiety response to each nutrient using a quadratic function, but this proved tedious and inaccurate, especially at the extremes.  Instead, I tried a more straightforward two-way linear regression, which provided a closer correlation with the data. 

The chart below shows the simplified two-way linear regression for protein (%).  Note the peak at the bliss point aligns with the RStudio GamFIT protein chart. 

Using the two-way regression model, we can more accurately predict a person’s daily energy intake based on the concentration of each nutrient in their diet. 

Single vs. 2-Way Linear Regression Results

The table below shows the correlation coefficient (R2) between the predicted and actual calorie intake for each nutrient when considered separately using the simple linear regression and the two-way linear regression. 

Nutrient	one-way R2	two-way R2
protein %	-21.9%	23.7%
potassium	-17.9%	22.0%
MUFA	19.3%	21.3%
riboflavin (B2)	-10.5%	19.2%
calcium	-13.6%	16.7%
sugar	10.5%	15.4%
vitamin C	-10.3%	15.0%
iron	-12.5%	14.0%
sodium	-8.5%	11.7%
fat	3.3%	7.4%
weight	-1.0%	3.1%

With the simple linear approach, we see that a higher concentration of protein, potassium, B2, calcium, vitamin C, and sodium, as well as a lower energy density, all align with a lower calorie intake.  Meanwhile, a higher concentration of fat, MUFA, and sugar per calorie aligns with eating more. 

However, in the far right column, we see that the two-way regression provides a significantly improved accuracy. 

Satiety Algorithm Weightings

Once I had modelled each of the essential nutrients, I created a combined satiety formula to predict energy intake.  This approach essentially uses all the simplified two-way regression charts stacked on top of one another to model our complex appetite.   

The solver function in Excel was then used to optimise the weightings to maximise alignment with the actual measured energy intake in the data.  The solver function runs numerous scenarios to identify the weightings for the individual nutrients that give the highest correlation with the measured energy intake. 

Nutrients that did not improve the accuracy were removed from the analysis, creating a shortlist of nutrients that, based on the data, we appear to have a statistically significant specific appetite for in our modern food system.

The chart below shows the weightings of each of the statistically significant satiety factors used in the final satiety algorithm.

With these weightings, we achieved an overall R2 correlation coefficient of 39.7%, which is a significant improvement over and above any single satiety factor (note: protein linear R2 = 21.9%) and multivariate linear regression (which had an R2 = 28.8%). 

That is, without knowing anything about a person’s weight, age, gender, muscle mass or activity, we can explain nearly 40% of the variation in their energy intake based on what they eat. 

We then convert the predicted calorie intake to a satiety score from 0 to 100%. The chart below shows the distribution of satiety scores in the combined dataset, with an average of 39.1%.

The chart below shows the relationship between the satiety score and energy intake based on the data, demonstrating that we have a direct relationship between the satiety score and energy intake across this dataset. 

Why Are These Nutrients Statistically Significant Satiety Factors? 

This analysis suggests that rather than merely protein leverage, there is a nutrient leverage effect at play.  When we consider a section of other nutrients, we can more accurately predict energy intake based on someone’s dietary choices. 

While we don’t have a robust conscious appetite for some of these nutrients, the data suggests they appear to play a subtle role in appetite when they are severely lacking or at very high concentrations. 

While limited human studies have been undertaken to demonstrate a specific appetite for some of the shortlisted nutrients (e.g. vitamin B2 and vitamin C), we do know that the concentrated form of most of these nutrients has a strong, unappealing taste, thus triggering sensory-specific satiety. 

It’s also worth noting that potassium and calcium are both nutrients of public health concern, according to the USDA, due to their low intake in the US.  Iron is also the number 1 nutrient deficiency internationally, according to the WHO, particularly in developing nations where there is minimal access to animal-based foods.  So, it’s logical that the statistically significant satiety factors would align with nutrients that are lacking in our modern food system.

However, the reason that some nutrients play a more prominent role in the weighted satiety algorithm could be due to the fact that there is higher quality data for them because they are mandated to be shown in the nutrition facts label by the FDA.  Meanwhile, other micronutrients are less frequently measured, so the data quality is poorer.  Thus, a larger weighting is given to the nutrients for which there is better data quality.  If we had higher-quality data for all the micronutrients, other nutrients would likely become statistically significant in the analysis. 

I hope the preliminary satiety analysis generates more interest for others who have the resources to undertake controlled studies to investigate this further.  However, it may be some time before this occurs due to various reasons, such as the large amount of data that would be required and the previously mentioned ethical issues with intentionally inducing nutrient deficiencies in humans.  Regardless, we can still use the data available to guide our nutritional decisions to increase satiety and nutrient density. 

Application of the Satiety Per Calorie Algorithm

There are a number of exciting potential applications for a quantified satiety per calorie algorithm.

Accurately Defining Ultra-processed Foods

Most people understand that they should prioritise fresh whole foods, like seafood, vegetables, meat and fruit, which are highlighted by the satiety algorithm.  The problem that most people perceive is that they feel ‘addicted’ to foods and can’t stop eating.   

Because our satiety algorithm focuses explicitly on identifying foods that hit multiple bliss points, it could be used as a more robust, quantified approach in place of NOVA Class 4 ultra-processed foods.   The satiety algorithm empowers users to identify foods that are designed to be eaten all day, every day, to maximise intake, whether made in a factory or the kitchen. 

It’s interesting to note that the sweeter treats have a slightly higher satiety score than savoury foods like burgers and pizza.  It’s important to remember that the satiety algorithm is calibrated using daily total energy intake, not just short-term cravings or hedonics.

Sweet, hedonic treats like milk chocolate may be more seductive for someone who has eaten relatively healthy meals during the day.  However, because they contain a lot of sugar, we would tire of them somewhat if that’s all we had to eat all day, every day because they would eventually exceed our capacity for sugar and trigger sensory-specific satiety.  In contrast, lower-satiety foods like burgers and pizza hit our bliss points more effectively without triggering sensory-specific satiety with any individual nutrient.  

High Satiety Foods

The most straightforward entry point to understanding how the satiety algorithm works is to look at the satiety score of foods.  People can use a satiety per calorie scoring system to compare one food to another and make more informed choices. 

For details on the satiety scores for various foods, including simple infographics, see:

• The Most Satiating Foods
• High Satiety Seafood
• Filling Fruit
• High Satiety Vegetables
• High Satiety Dairy & Egg Products
• Meats to Keep You Satisfied
• Beans, Legumes & Nuts
• Low-Satiety “Addictive” Foods

For more details, check out our interactive chart showing the satiety score of 750 popular foods. 

High Satiety NutriBooster Recipes

However, we usually don’t eat individual foods.  We combine them into meals.  So, we have included the satiety score in our database of 1750 NutriBooster recipes, which we use in our Macros Masterclass and Micros Masterclass.

Daily Food Intake

Because our satiety score is calibrated using daily energy intake data, the ideal way to implement it is to use the satiety algorithm to calculate a daily satiety score, which the user can work to incrementally increase by swapping out lower satiety foods for higher satiety options. 

In our Macros Masterclass, we have our Optimisers track their current food for a week and then show them which foods and meals to prioritise vs reduce to incrementally increase their satiety score.  This incremental approach to improving satiety forms new habits.  It avoids radical dietary shifts that can lead to excessive hunger or reverting to old habits ‘once the diet is over’. 

Summary

• Analysis of 654,476 days of food intake data shows a craving/sensory-specific satiety response curve for each essential nutrient, with a distinct bliss point once the minimum intake has been reached. 
• Combining these curves using a two-way linear regression analysis allows us to estimate calorie intake and thus develop a satiety per calorie score that can be applied to other foods, meals, food groups or daily food intakes.
• While protein leverage is the dominant satiety factor and energy density is useful, combining these two approaches to satiety, along with other nutrients, enables us to more accurately predict how much we will eat and thus actively manage satiety. 
• This data-driven approach to satiety has the potential to help users make better food choices and decrease their daily energy consumption by proactively satisfying cravings with less energy. 
• This nutrient-focused analysis of our satiety response to food generates a number of hypotheses (e.g., nutrient leverage) that can be tested by others in future studies. 

Appendix A – Common Questions and Objections

This appendix addresses some common questions and objections to the satiety scoring system.  If you have anything further that you would like to see discussed, feel free to leave a question or suggestion in the comments below. 

Individual Foods vs Daily Intake

Unfortunately, satiety per calorie is not intuitive.  When many people first look at the satiety scores of individual foods, they often find it absurd that foods like mushrooms might be more satisfying than eggs or steak, which they (subjectively) find satiating (without regard for total daily calorie intake). 

But the reality is that no one eats only spinach or asparagus all day, every day, so considering foods alone has limited value.  Thus, the messaging around the satiety scores needs to emphasise that the scoring is based on daily energy consumption and that higher satiety foods can be valuable additions to one’s daily diet. 

Nutrient Density

The satiety analysis highlights that the foods we tend to eat the most are not nutrient-poor, empty calories.  Instead, the lowest-satiety foods, like burgers and pizza, contain just enough of the nutrients we need to hit our bliss points without triggering sensory-specific satiety. 

In contrast, if all we had to eat was nutrient-poor foods like butter, white rice or sugar for days on end, we would quickly tire of them and eat less.  Sugar alone would soon exceed the storage capacity in our blood, liver and muscles, and we would soon tire of these foods. 

As shown in the snip below from our interactive food search tool, some nutrient-poor foods have a moderate satiety score.   We could become malnourished by eating some high-satiety foods because they are bland and nutrient-poor. 

In contrast, nutrient-rich high-satiety foods have fuller, stronger taste because they contain more of the nutrients we require.   Hence, it’s critical to interpret the satiety score in the context of daily intake with an eye on nutrient density. 

Calories & Thermic Effect of Food

Some people object to using calories as the denominator of the satiety scoring system (i.e. satiety per calorie) due to the complexity of accurately measuring energy balance. 

Similar to tracking nutrients in food, there are multiple inaccuracies in our calculation of the potential chemical energy in food and how it converts to kinetic energy and stored energy in our bodies. 

However, these inaccuracies and complexities in human metabolism that we can’t yet quantify with 100% precision do not stop the satiety score from being helpful or directionally useful. 

The thermic effect of food (or specific dynamic action) is a crucial consideration here.  That is, we lose more energy metabolising protein (20 to 35%) than carbohydrates and fat (3 to 15%). 

While accounting for the thermic effect of food to estimate available energy would add another layer of complexity, possibly making the system more accurate, it would bias the system even more towards protein-rich, higher-fibre foods that already have a high satiety score and hence might be redundant. 

What About The ‘Bad Things’ In Food? 

As noted earlier, many people define their diet based on avoiding ‘bad things’ in food (e.g. plant-based foods, animal-based foods, toxins, oxalates, saturated fat, PUFA, seed oils, etc.), which, mainly due to a lack of data, are not included in our satiety per calorie model.

However, this is not considered to be a significant concern, given that higher satiety, nutrient-dense foods tend to automatically reduce these ‘bad things.’ 

• Higher satiety foods automatically prioritise minimally processed whole foods and exclude heavily processed packaged foods that contain seed oils, flavours, and colouring. 
• Nutritious, high-satiety foods also reduce nutrient-poor refined grains, which provide the majority of the antinutrients in the modern diet that inhibit the absorption of other nutrients.
• Some nutritious, high-satiety foods contain oxalates, but they also include plenty of calcium to keep the calcium:oxalate ratio high enough to ensure adequate clearance of oxalates. 
• As shown in the chart below, a higher concretion of any of the fats aligns with eating more.  Still, only MUFA and total fat % made it into the shortlist of nutrients included in the satiety algorithm because saturated fat tends to come with protein, which is a crucial satiety factor.  PUFA is not a large part of our diet and, hence, is not a significant satiety factor.  

Blood Glucose and Insulin

Some have objected to the fact that the satiety scoring doesn’t intentionally bias against carbohydrates, which is relevant for people managing diabetes or those who prefer a lower carbohydrate diet.

To assist people who currently have dysregulated blood glucose levels, we have added a filter for carbohydrates per serving and carbohydrate % to the food search tool and recipe database.  This allows users to filter based on carbohydrate preference or tolerance.  In our Macros Masterclass, we recommend Optimisers reduce their carbohydrate intake if they see their glucose typically rise by more than 30 mg/dL or 1.6 mmol/L after eating. 

As a general rule, foods with a higher satiety score are lower in non-fibre carbohydrates and contain more protein and fibre.  As shown in the satiety response chart below, moving from the 48% carbohydrate bliss point to around 20% results in a significant reduction in energy intake. 

While it is possible to create a high-satiety diet using very high-carb, low-protein foods (e.g., fruit and vegetables), the data suggests that this approach is less likely to be less nutritious than a protein-focused, high-satiety approach. 

A lower carbohydrate approach will also lower the magnitude of the rise of the postprandial insulin response to food.  However, any version of a higher satiety diet that leads to weight loss (as intended) will improve insulin sensitivity, lower basal insulin and lower total insulin levels across the day. 

What if I’m Allergic to Certain Foods or Prefer a Certain Dietary Approach?

We have added a drop-down filter to our food search tool to cater to people who prefer a specific dietary approach (e.g., plant-based, fruitarian, vegetarian, pescatarian, or carnivore).  

We have also divided the NutriBooster recipes that we use in our Macros Masterclass and Micros Masterclass into the following categories to enable Optimisers to cater to their own preferences and allergies:

• Meat-based,
• Pescitarian,
• Dairy-free,
• Vegetarian,
• Lacto vegetarian,
• Plant-based,
• Egg-free,
• Dairy & egg free,
• Autoimmune Paleo (AIP),
• Low FODMAP,
• Low oxalate and
• Low histamine.

What About Fibre?

Fibre, particularly the fibre:carbohydrate ratio, is a valuable marker of minimally processed carbohydrate-containing foods.  When considered alone, fibre can be a helpful satiety factor.  However, fibre is not a statistically significant satiety factor when the vitamins and minerals are also considered. 

What About Nutrient Bioavailability?

The satiety weightings have been calculated based on the available data, which does not account for bioavailability.  However, a separate analysis of the low-carb vs low-fat data has shown that, in line with other studies, some nutrients, like iron (shown below), B2, sodium, calcium and vitamin C, have a marginally lower bliss point for the low-carb data subset while the bliss point is the same for many other nutrients like potassium and vitamins B5, B6 and E.

The lower bliss point is likely due to a higher bioavailability of some nutrients from meat, seafood, and dairy compared to plant-based foods.  This will be discussed in a future article and could be included in future iterations of the satiety algorithm.  Still, while it would make the algorithm more complex and less transparent, it is unlikely to change the directionality of the recommendations.

There is No One Size Fits All Satiety Algorithm

Our latest iteration of the satiety algorithm is calibrated using 654,476 days of data from people following a range of diets all over the world.  It shows the weightings of the nutrients that will help most people improve their satiety most of the time.  But when we dissect the data into low-carb vs. low-fat subsets, we see that the weightings change depending on which nutrients are lacking in that subset of the data.  

While a little bit more work is needed, this can be addressed on an individual basis by prioritising foods that contain the nutrients that a person is currently missing from their diet to ensure they are exceeding the bliss point concentration and moving towards the optimal nutrient intake for those nutrients. 

The nutrient fingerprint shows how we manage this in the Micros Masterclass, highlighting foods and meals to prioritise and increase nutrient density and satiety.  However, this approach requires tracking for a few weeks, which is a bit more effort. 

While the Micros Masterclass is excellent for people wanting to fully optimise their diet at the micronutrient level, the high-satiety foods lists and NutriBooster recipes will likely be preferred by most people, at least as a starting point. 

Further Refinement

As noted above, we have used a two-way linear regression to model the inverted U-shape craving/satiety response to each nutrient.  We will continue to investigate ways to model this complex relationship more accurately to improve the correlation with actual energy intake further.

As you can see from this article, our satiety scoring system has been a work in progress for six years.  We plan to continue reviewing and updating the algorithm at least annually as more data becomes available (e.g., from people taking our Nutrient Clarity Challenge to identify their priority nutrients and ongoing updates to the FAO/WHO dataset).

Finally, while we have been successfully implementing and refining this approach with our Optimiser Community in our Macros Masterclass and Micros Masterclass for the past five years, it would be exciting to run a prospective randomised control to demonstrate the results of this quantified approach to satiety. 

If you are a researcher who also believes that a nutrient-focussed approach to satiety might have an exciting future and you have a fat stack of cash available to fund a study, don’t hesitate to reach out!  

Appendix B – Satiety Response Curves

For completeness, I’ve included the individual satiety response curves for the statistically significant satiety factors included in the algorithm. 

Protein %

Fat

Potassium

Monounsaturated Fat

Riboflavin (B2)

Calcium

Sugar

Vitamin C

Iron

Sodium

Energy Density

More on Satiety

• High satiety foods
• Low satiety foods
• Sensory Specific Satiety
• Food Satiety Index
• Sensory specific satiety
• Satiety impact of minerals
• Satiety impact of vitamins
• Satiety impact of amino acids
• How to increase GLP-1 naturally
• How to stop food noise