(Dis)counting calories — why you shouldn’t read calorie information on food labels
The year 1824 marks the birth of the word “calorie”, as the amount of energy necessary to raise the temperature of one kilogram of water by one degree Celsius. Although often attributed to the father of modern chemistry, Antoine Lavoisier (1743–1794), it was probably conceived by Nicolas Clément, another French scientist.
The calorie became the standard unit of energy in food science, contrary to the rest of modern science. This is mostly due to an American chemist, Wilbur Atwater, using it in his extensive research on the energy value of food, around the end of the 19th century. If you have the habit of reading, and believing, calorie information on food labels, you might want to get acquainted with his work.
Atwater expanded on methods used by his German advisor, Max Ruber, to quantify energy provided by different foods. These included burning food in a calorimeter — a lab equipment for measuring the quantity of energy stored in a substance, and feeding people inside special isolated chambers. He then calculated the famous values for the calorie content of macronutrients - 4 cal/g of protein, 4 cal /g of carbohydrate and 9 cal/g of fat. Notice that I used the word “calculated”, not “measured”. Why?
Determining the energy provided by each type of macronutrient presents three basic challenges.
First, there is a lot of variation for each macronutrient. Glucose and starch are both carbohydrates, but have different energy content. Starch in whole wheat behaves differently that in refined wheat. One single value cannot properly express the energy content of a type of macronutrient in all situations.
Second, food processing in our bodies is much more complex, and less efficient, than what happens in a calorimeter. Values obtained in lab equipment are only an upper limit in ideal conditions.
Third, quantifying the energy supplied by each type of macronutrient depends on being able to determine food composition. But measuring the content of each type of macronutrient in a given food item is not as straightforward as it might seem.
USDA Agriculture Handbook No 74 — Energy value of foods: basis and derivation, or the “Handbook” (as I will refer to it in the rest of this text), discusses these issues as they apply to food energy data obtained until 1955. It remains as the main source of data for nutrition labels in the US, although its limitations have been largely ignored outside research labs.
The variability in forms of each macronutrient was handled by the common, but often misleading, method of averaging. Atwater first came up with values for the gross energy content of protein, fat and carbohydrate. When used with the experimentally determined composition of a food item, they yielded energy values close to what was obtained by burning it in a calorimeter. He showed that the difference between the energy values obtained by the two approaches was only 0,3%, on average (table 5 of the Handbook).
But the average says nothing about data variability, and the original data shows, for example, that Atwater’s method overestimated energy content of cooked vegetables by 5.4% and underestimated it for fresh fruit by 6.1%. Also, the universe of food samples he used might be representative of a late 1800s diet, but not exhaustive by modern standards. Atwater analyzed 276 samples of 25 types of food, but one third of the samples were either beef or milk. Not that impressive. And we haven’t even started to consider what happens to food in the body.
Since part of ingested food ends up as faeces and urine instead of energy, it is obvious that the amount of energy our body gets from food must be lower than the gross calorie content obtained as described above. Atwater therefore came up with correction factors for the “availability” of each macronutrient. And the curse of the average comes up again. For protein, he obtained an average of 92% from values that ranged from 83 to 97%. Values for fat ranged from 90 to 95% and Atwater settled for 95%, based on a typical diet at the time. Applying these correction factors to the previously determined gross energy content for each macronutrient, Atwater reached the famous values of 4–4–9 cal/g for carbohydrates, protein and fat.
All these numbers depended, however, on the determination of the content of protein, fat and carbohydrate in different foods. These involved many challenges, some of which remain to this day. The publication Food energy — methods of analysis and conversion factors, a report from a FAO Workshop, summarized the situation in 2002.
Because of the complexity of more precise methods, protein content is commonly derived from the total nitrogen content of the food. That involves two big assumptions that are not really true. First, non-protein nitrogen (such as in free amino acids, nucleotides, creatine and choline) is ignored. By a happy coincidence, most foods that contain a lot of non-protein nitrogen have a low total nitrogen content, so in most cases this error is not very significant for the total calorie count. A conversion factor is then used to calculate the weight of protein in the sample from its nitrogen content. The factor generally used is 6.25, once again an average, since conversion factors for individual proteins range from 5.26 to 7.19.
Methods for determining fat content are the most reliable. On the other hand, carbohydrate content is generally determined by difference (100-%protein-%fat-%water-%ash-%alcohol), so it inherits the errors from all other values. As a side note, Atwater was the first to estimate alcohol’s energetic value, which did not fit well into the conservative culture at the time.
Another problem for carbohydrate determination is that not all of it can be digested by the human body. Dietary fiber is composed of carbohydrates, most (but not all) of which will come out as waste without being converted to energy. There are modern methods to directly determine the available carbohydrate content of a given food item, but they are not used in most cases.
Despite its limitations, Atwater’s work was groundbreaking and his assumptions made sense in the context he worked in. The problem is taking Atwater’s results at face value and using his numbers for a different purpose than what they were designed for. His factors were averages obtained for use in evaluation of diets, and based on food habits of his time. Do you believe your diet is similar to a typical diet in 1900? Probably not. But much worse is the later practice of applying Atwater’s factors to individual foods. As the Handbook observes,
The 4,9,4 factors later came into widespread usage in estimating calories of food and not only were applied to the total amount of fat, protein and carbohydrate (…) of a mixed diet as Atwater and Bryant had originally intended but also were used in assessing the fuel value of individual foods.
A lot of work was done in the first half of the 20th century to adapt Atwater’s factors to specific foods. This data is summarized in Table 13 of the Handbook, which shows some surprising results. Many people would be happy to know, for example, that the values obtained for chocolate were 1.83 cal/g of protein, 8.37 cal/g of fat and 1.33 cal/g of carbohydrate. As an exercise, try finding one food group that matches that 4–4–9 cal/g model…
Why do these differences exist? Atwater’s average factors do not work for individual food items because “the compositional and structural features of foods can drastically change the bioavailability of nutrients”. And processing further affects these features, which is particularly relevant as today more that 3/4 of calories in US households come from processed foods.
This problem was acknowledged from the beginning. The Handbook concludes that despite their limitations, Atwater’s factors were the best that could be done at the time. Until better data was obtained, the specific factors of Table 13 should be used:
It is recognized that some of the physiological fuel factors for food groups and individual foods developed […] are based on a limited amount of data and that factors for food groups may not always be equally suitable for individual foods within the group. Also revisions are anticipated as more information becomes available […] Moreover it is realized that there are problems with direct bearing on the digestibility of protein, fat, and carbohydrate that have not been satisfactorily resolved at this time. Although all of the calorie factors may not be entirely suitable as a result of the various limitations existent in the basic data, nevertheless when they were applied to the nutrients in food fed alone or in various combinations, the estimated total available energy of the food was always in excellent agreement with the value determined by use of the bomb calorimeter.
In view of the agreement noted and until more basic information becomes available, the modification of the method of Atwater […] as proposed in the present publication […] seems the most satisfactory procedure to use; the calorie factors presented in table 13 are recommended for calculating the total available energy value of foods until there is basis for further revision or refinement of the factors.
Note the main argument for the use of calorie factors in the calculation of energy content in foods was, besides lack of better information, that “the estimated total available energy of the food was always in excellent agreement with the value determined by use of the bomb calorimeter.” In other words, we could predict what a lab equipment would tell us about the food, not what would actually happen to it in the body.
Even for diets, more recent data show that the use of Atwater factors is somewhat questionable. That is not surprising, as they are average values calculated for a given 19th century diet. For diets of different composition, we would expect the averages to be different. One study, for example, concluded that the Atwater general factors were good enough for the typical British diet of the 1970s, if 3.75 cal/g was used for carbohydrates. Remember that if you are trying to lose weight while traveling with Dr. Who.
Overestimation of calorie content by the use of Atwater factors can be as high as 18% for diets with high fiber content. These include those with more plant-based foods, like vegan and Mediterranean diets.
This error may contribute to the sometimes confusing results that come out of studies that try to determine the best diet for losing weight. It is hard to compare the effects of different diets of equal calorie content if they are not really equal.
Anyway, you are probably still thinking about that lower calorie content of chocolate. So when you pick up a bar of chocolate at the market, where does the calorie information on the label come from? The general factors or the specific factors determined for chocolate? Well, it depends… In Europe, Atwater’s general factors are used, with corrections for different types of carbohydrates. Not the best method, but at least you know how it was done. In the U.S., CFR Title 21, section 101.9(c)(1) regulates nutritional information presented in food labels. It states that calorie content may be calculated by any of the following methods: the Atwater general factors, the general factors with an adaptation for non-digestible carbohydrate, specific food factors or calorimeter data with an adaptation for protein content. And there is no obligation to state which method was used. Comparing calorie counts when choosing what to eat is, therefore, a complete waste of time.
It is therefore not surprising that the FAO Workshop on food energy concluded that there was a “major need for rationalization and harmonization of methods of food analysis and energy conversion factors”. The effort necessary to correct the situation was obvious, so that “The participants at the workshop recognized that this is no small task, but believe it is a task that can be accomplished gradually over a number of years, if scientists and regulatory authorities have the will and the willingness to work together to that end.”
Even if we correctly determined the energy content in food, we would not be measuring the actual amount of energy available to us, as energy from food does not power our body directly. It is used to produce ATP (adenosine triphosphate), the molecule that serves as body fuel, through a number of metabolic pathways. A proper measurement of what results from that complex process is called “net metabolyzable energy”.
This means that even worse than using calorie information to make diet choices is comparing them to estimates of calories burned during exercise. Even we if disregard the innacuracy in these estimates (a subject for another time), the best, and rarely used, methods to calculate calorie content for food labeling give only the “metabolizable energy”. That is “at best an approximate surrogate for net metabolizable energy”, which is still one step away from the energy burned in exercise.
Among other issues, this practice leads to the argument that exercising is not relevant for weight loss because the number of calories burned is equivalent to a small quantity of ingested food, so it is more effective to eat less. It may well be, but not because of this crude calorie math. Even if these comparisons are promoted by reputable sites, don’t plan your meals and workout believing that one hour of running is equivalent to one Big Mac with fries.
It is important to note that inaccuracies in measurement are a basic fact of science. It is impossible to know the exact true value of any quantity in the real world. Proper scientific measurements are always accompanied by an uncertainty estimate, which defines a range of values between which we believe the true value lies, with a given degree of confidence.
The key point, then, is to always use values that were measured with an accuracy adequate to what they will be used for. The precision of a simple ruler is enough if you want to decide if a desk will fit in your bedroom. It is not enough if you want to do quality control on auto parts. The problem with the widespread use of calorie counting is that the numbers are being used in a way not justified by the accuracy in their determination. That leads to a lot of wasted time, bad decisions and personal frustration.
If you want a healthy life, pretend it’s the good old days when people enjoyed food instead of doing math on it. Eat more real food and less processed junk food. Don’t eat too much. Have fun being as active as possible. And forget you ever heard of calories.