🔥How TDEE is calculated, in full

The shape of the calculation

TDEE is built in two steps that are not equally well founded, and almost every disagreement between calculators comes from failing to distinguish them. Step one estimates the energy your body uses at rest, from a regression fitted to measured resting metabolic rates. Step two multiplies that figure by an activity factor to account for everything you did on top of lying still. Step one has a paper behind it. Step two does not.

Keeping the two separate is not pedantry. It tells you which half of the answer you can defend, which half you chose, and where to look when two calculators hand you numbers 600 kcal apart.

Step one: Mifflin-St Jeor, term by term

The equation Mifflin, St Jeor and colleagues published in 1990 is: resting energy expenditure in kcal per day equals 10 times weight in kilograms, plus 6.25 times height in centimetres, minus 5 times age in years, then plus 5 for men or minus 161 for women.

Each coefficient is a fitted slope, not a mechanism. The weight term is the largest because lean tissue is metabolically active and heavier bodies generally carry more of it. The age term is negative and small: five kcal a year, so thirty years costs 150 kcal a day at the same height and weight. The two constants are what makes the male and female versions of the line meet the axis in different places; they are properties of a regression, not a statement about anyone.

For an 80 kg, 180 cm, 30-year-old man that is 800 + 1,125 - 150 + 5, which is 1,780 kcal a day. The same measurements with the female constant give 1,614. That difference of 166 kcal is the two constants, 5 and -161, and nothing else.

The two equations that compete with it

The Harris-Benedict equations date from 1919 and were refitted by Roza and Shizgal in 1984 using the original 239 subjects plus 98 more. For the same reference body the revision gives 1,854 kcal. Roza and Shizgal state in their own paper that the equations estimate an individual's resting energy expenditure to a precision of about 14 percent, which is a wider band than the one usually quoted for Mifflin-St Jeor.

Cunningham's 1991 equation - the one textbooks carry as Katch-McArdle - takes a different input entirely: resting metabolic rate equals 370 plus 21.6 times fat-free mass in kilograms. It ignores height, age and sex, on the reasoning that what they were standing in for all along was lean tissue. Given a Boer-derived lean mass of 61.42 kg for the reference body it returns 1,697 kcal.

So the three land at 1,697, 1,780 and 1,854 - a spread of 157 kcal. Notice the conditional attached to the third: it needs a lean body mass figure, and if that figure is itself an estimate, this equation multiplies its error by 21.6. Fed Hume's estimate of 57.79 kg instead, it returns 1,618 and the spread widens to 236 kcal. An equation is only as good as its worst input.

Step two: the multiplier nobody published

Every online calculator offers the same ladder: 1.2 for sedentary, 1.375 lightly active, 1.55 moderately active, 1.725 very active, 1.9 extra active. No primary source publishes those five numbers. They are a field convention, repeated until they acquired the appearance of a finding, and this site marks them as a convention in its formula pack rather than dressing them up as one.

What is published is the physical activity level framework in the 2004 FAO/WHO/UNU report on human energy requirements. It defines PAL as total energy expenditure divided by basal metabolic rate - the same ratio the multiplier is standing in for - and documents three lifestyle bands for adults: 1.40 to 1.69 for a sedentary or light activity lifestyle, 1.70 to 1.99 for an active or moderately active one, and 2.00 to 2.40 for a vigorously active one.

Lining the ladder up against those bands produces one genuinely useful fact. The bottom two rungs, 1.2 and 1.375, fall below the lowest band FAO/WHO/UNU documents at all. This site says so on the page rather than silently offering them as ordinary choices.

Who the equations were fitted on

Every equation on this page is a line drawn through a group of people, and the group is part of the result. Mifflin, St Jeor and colleagues fitted theirs on healthy adults. Roza and Shizgal refitted Harris-Benedict from the 239 subjects behind the 1919 equations plus 98 further subjects published by Benedict. Cunningham's is a synthesis of previously published data rather than a fresh cohort. Frankenfield's 2005 review, which produced the 10 percent criterion, covered healthy adults both nonobese and obese.

This is why the phrase "how accurate is this equation" is slightly the wrong question. An equation is accurate to the extent that you resemble the people it was fitted on, and the review can tell you how often it lands within 10 percent across a population but not whether you are one of the people it lands near.

The sex input deserves the same reading. Every one of these equations was fitted on two groups, so the input is asking which of two fitted lines to use rather than making a claim about anybody. This site says so in the schema for the field rather than leaving the reader to infer it, because a coefficient of -161 kcal is a property of a regression and nothing else.

The error band, and why it is wider than it looks

Frankenfield's 2005 systematic review compared resting-metabolic-rate equations by asking how often each landed within 10 percent of a measured value. Mifflin-St Jeor did so for more people than any other equation reviewed. That is the strongest available endorsement of it, and it necessarily also means the equation missed by more than 10 percent for some of the people in those studies.

Applying that band to the reference body turns 1,780 kcal into 1,602 to 1,958. Carry it through the 1.55 multiplier and a TDEE that reads 2,759 becomes 2,483 to 3,035 - a span of 552 kcal a day. That is more than three times the disagreement between the three resting equations, which is the disagreement people actually argue about.

And the true band is wider still, because the multiplier contributes error that has never been quantified. There is no honest way to put a number on it, so this site does not invent one; it says the multiplier is a convention and leaves the sentence there.

Checking the estimate against reality

The way out of this is not a better equation. It is treating the estimate as a hypothesis and the scale as the instrument. Hold intake roughly steady for two or three weeks, track weight as a trend rather than a daily reading, and the direction of travel tells you more about your own expenditure than any regression fitted on other people.

Expect that to be slow, because it is. Hall and colleagues published a rule of thumb in The Lancet in 2011: for an average overweight adult, a sustained change of about 100 kJ a day in intake corresponds to an eventual bodyweight change of about 1 kg - roughly 10 kcal a day per pound - with about half of that change reached in about a year and about 95 percent of it in about three. Weight responds to a sustained change on a timescale of years, not the timescale of a calculator.

What TDEE is not

It is not a measurement. Total daily energy expenditure is measured with doubly labelled water, and resting expenditure with indirect calorimetry. A web form is doing neither of those things; it is evaluating a line fitted through other people.

It is not stable day to day either. Expenditure moves with sleep, ambient temperature, what you did yesterday, and how much you move without meaning to - a category with its own name, NEAT, and a large between-person range. A single figure to the calorie implies a constancy that the underlying quantity does not have.

And it is not a target. TDEE and maintenance calories are the same quantity under two names: the intake that leaves weight where it is. What anyone should do with that figure is outside what arithmetic can answer, and this site stops where the arithmetic does.