BMR and TDEE Formulas: A Century of Equations, How Accurate They Are, and Why the 3,500-Calorie Rule Misleads

Type your height and weight into a fitness app and it hands back something like “maintenance: 2,340 calories.” The number looks precise. It is actually the product of two steps: a regression equation whose lineage goes back to a laboratory in 1918, and an activity multiplier whose origin is hard to pin down. This article traces both, looks at how far off they can be, and explains why the familiar “cut 500 a day, lose a pound a week” promise breaks down after the first couple of months.

Measuring metabolism before there were equations

Basal metabolic rate (BMR) is the energy your body uses lying awake and still after an overnight fast: the cost of a beating heart, breathing, keeping warm and running the brain, liver and kidneys. In the early twentieth century Francis G. Benedict ran the Nutrition Laboratory of the Carnegie Institution of Washington, where subjects breathed into equipment that measured oxygen consumed and carbon dioxide produced. From those gas volumes you can calculate heat production. The technique is called indirect calorimetry, and it is still the reference method today.

In 1918 Benedict and the biometrician J. Arthur Harris published “A Biometric Study of Human Basal Metabolism” in the Proceedings of the National Academy of Sciences, followed by a full monograph in 1919. They fitted separate equations for men and women that predicted measured BMR from weight, height and age, using data from 239 people. What made the result last was its convenience: three numbers anyone could collect, plugged into a formula anyone could compute. For most of the century it was the default in hospitals and textbooks.

A refit, then a replacement

In 1984 A. M. Roza and H. M. Shizgal reworked the Harris-Benedict equations. They added 98 subjects from a later Benedict series that covered a wider age range and refitted the coefficients. The new equations were very close to the old ones, and Roza and Shizgal reported that they estimated resting energy expenditure in normally nourished people with a precision of about ±14%. They also found them unreliable in malnourished patients. Their version, with coefficients like 88.362 + 13.397 × weight for men, is the “revised Harris-Benedict” in most calculators.

It helps to translate ±14% into calories. If the equation says your BMR is 1,500 kcal, the true value could be anywhere from 1,290 to 1,710. Multiply by an activity factor of 1.5 and the range of plausible daily needs is more than 600 kcal wide.

In 1990 a group at the University of Nevada School of Medicine led by M. D. Mifflin and S. T. St Jeor built a new equation from scratch. They measured resting energy expenditure in 498 healthy adults aged 19 to 78. Unlike Benedict’s sample, almost half of theirs (234 people) had obesity. After simplifying, the equations became:

• Men: 10 × weight (kg) + 6.25 × height (cm) − 5 × age + 5
• Women: the same, with −161 instead of +5

In their data the 1919 Harris-Benedict equations overestimated measured values by 5% on average. The paper does not say why, but one plausible reading is simply that an equation fitted to people in the 1910s drifts when applied to a heavier, more sedentary population seventy years later.

Lean mass does most of the work

A lesser-known result sits in the same Mifflin paper. The equation using weight, height, age and sex explained 71% of the variation in measured energy expenditure (R² = 0.71). Fat-free mass alone explained 64% (REE = 19.7 × FFM + 413). Weight alone explained 56%.

A decade earlier, J. J. Cunningham had reanalysed 223 subjects from the original Harris-Benedict data and found that lean body mass was the single best predictor of BMR. His equation was 500 + 22 × LBM, and he concluded that sex and age added little once lean mass was known. Put simply, much of the reason men burn more than women, and older people less than younger ones, is the difference in muscle and organ mass. The Katch-McArdle equation used in many calculators (370 + 21.6 × lean mass) rests on the same idea.

The catch is that lean mass requires a body fat percentage. A tape-measure or BMI-based estimate from a body fat calculator carries an error of several percentage points, and that error passes straight into the BMR. For people with roughly average muscle, a lean-mass equation often lands close to Mifflin anyway. It earns its keep at the extremes: very muscular people, or people with little muscle for their weight.

What the evidence says about accuracy

In 2005 an evidence analysis team working for the American Dietetic Association, led by D. Frankenfield, published a systematic review of prediction equations for resting metabolic rate in healthy adults. The practical question behind it was whether dietitians could rely on equations or needed to measure metabolism directly. Crucially, the team only included studies that reported errors for individuals, not just group averages, because that is how dietitians use the equations.

They identified four equations as the most used in clinical practice: Harris-Benedict, Mifflin-St Jeor, Owen and WHO/FAO/UNU. Their findings:

• Mifflin-St Jeor predicted resting metabolic rate within 10% of the measured value in more people than any other equation, in both people with and without obesity, and had the narrowest error range.
• Even so, noteworthy errors remained when it was applied to individuals.
• No validation work on individual errors was found for the WHO/FAO/UNU equation.
• Older adults and US-resident ethnic minorities were underrepresented in both developing and validating the equations, and the authors advised “a high level of suspicion” about accuracy for groups that are poorly represented.

So the best available equation is best in the sense that it is most often within about 10%. It does not know your BMR to the nearest ten calories, and for some people it will be further off than that.

Activity multipliers: the bigger source of error

To turn BMR into total daily energy expenditure (TDEE), calculators multiply by an activity factor. The same five values turn up almost everywhere: 1.2 for sedentary, 1.375 for light exercise, 1.55 for moderate, 1.725 for hard and 1.9 for very hard exercise plus physical work. It is surprisingly difficult to trace these to an original study that measured them, so they are best treated as a convention.

The FAO/WHO/UNU expert consultation of 2001 offers a documented comparison. It defines physical activity level (PAL) as TDEE divided by BMR and groups lifestyles like this:

Lifestyle	PAL
Sedentary or light activity	1.40–1.69
Active or moderately active	1.70–1.99
Vigorous or vigorously active	2.00–2.40

Its lowest band starts at 1.40, well above the calculator convention of 1.2 for sedentary people. At the same time, it is easy to pick an activity level above your real one, so the two errors sometimes cancel and sometimes stack.

Either way, the multiplier matters more than the equation. For a 30-year-old man who is 180 cm (5 ft 11 in) and 80 kg (176 lb), Mifflin gives a BMR of 1,780 kcal and revised Harris-Benedict gives 1,854, a gap of 74 kcal. Move the same man from “moderate” (1.55) to “hard” (1.725) and his estimated TDEE rises from 2,759 to 3,071 kcal, a jump of 312.

The most accurate multiplier is the one your own records produce. Log what you eat for two or three weeks and watch the trend in morning weight. The intake at which weight holds steady is your real TDEE. The output of a calorie calculator is a sensible first guess for that experiment, not its answer.

Why 3,500 calories does not equal a pound, over time

The rule says a pound of body fat holds about 3,500 kcal, so a daily deficit of 500 kcal removes a pound a week. As a statement about the energy stored in fat tissue, that is a fair approximation. The problem is treating it as a rate that holds indefinitely.

Kevin Hall of the US National Institutes of Health took this apart in a 2008 paper, calling it “one of the most pervasive weight loss rules.” Using a model of how the composition of lost weight depends on starting body fat, he showed that people with more body fat need a larger cumulative deficit per unit of weight lost, because more of what they lose is fat. Leaner people lose proportionally more lean tissue, which stores less energy, so a given deficit moves the scale further. In his analysis the rule roughly matched only for people with more than about 30 kg of initial body fat and overestimated the required deficit for everyone leaner.

The larger issue is that the deficit shrinks. As weight falls, it costs less energy to maintain and move a smaller body, so eating the same amount produces a smaller and smaller gap. In a 2011 Lancet paper, Hall and colleagues modelled these dynamics and proposed a replacement rule of thumb for adults:

• Every sustained change in intake of 100 kJ per day (about 24 kcal) eventually changes body weight by about 1 kg.
• About half of that change happens in the first year, and about 95% within three years.

They pointed out that the common advice that cutting about 2 MJ (roughly 500 kcal) a day produces steady loss of 0.5 kg a week ignores the body’s adaptation to its new weight. Applied to a 500 kcal daily cut, the two rules diverge sharply:

Rule	After one year	Long run
3,500 kcal per pound (7,700 kcal/kg)	about 52 lb / 23.7 kg	keeps going indefinitely
Hall’s rule of thumb	about 22 lb / 10 kg (half the total)	levels off near 46 lb / 21 kg

Those figures come from simple arithmetic on each rule (500 × 365 ÷ 7,700 for the first; 500 ÷ 24 for the second) and describe an average adult, not a promise for any individual.

In 2013 Diana Thomas and colleagues compared the 3,500-calorie rule’s predictions with the actual results of seven weight-loss experiments in which food intake was supervised or objectively measured. The rule grossly overestimated the weight people really lost. They also released free simulators, including one at the Pennington Biomedical Research Center, that model weight change dynamically.

In everyday terms: the first weeks of a diet often track the arithmetic reasonably well. Then, a couple of months in, comes the point where “I’m eating exactly the same and nothing is happening.” That is usually not a failure of willpower. Energy expenditure has come down to meet the new, lower weight, and the target needs recalculating.

Using the numbers well

• Start with Mifflin-St Jeor. It was the most often within 10% in the comparison studies. If you have a trustworthy body fat figure and your muscle mass is clearly above or below average, compare it with a lean-mass equation.
• Pick the lower activity level when unsure. Do not add your watch’s exercise calories on top of an activity multiplier; that counts the same workout twice.
• Calibrate against two or three weeks of data. Treat the calculated number as a hypothesis and your weight trend as the measurement. Adjust in steps of 100–200 kcal a day.
• Recalculate after every few kilograms lost. Slowing progress is expected, not a sign something is wrong.
• Look beyond the scale. A BMI calculator, waist measurements and a body fat estimate each have different blind spots; together they catch what one alone misses.