How accurate are BMI, BMR and TDEE estimates?

Good enough to start from, not precise enough to lean on. BMI is a screening number that can’t tell muscle from fat. BMR and TDEE come from equations fitted to groups of adults. For one person the resting estimate can miss by 10% or more in either direction, and the activity level you pick can move the daily total by several hundred kcal. Treat each number as a first guess to check against a few weeks of real life.

What do BMI, BMR and TDEE actually estimate?

BMI describes weight for height. BMR and TDEE estimate the energy you use at rest and across a whole day.

The Mifflin–St Jeor equation, this site’s default for BMR, was fitted to 498 healthy adults aged 19 to 78. Weight, height, age and sex explained 71% of the person-to-person differences in their measured resting expenditure ($R^2 = 0.71$); the other 29% is left unexplained. The same study found fat-free mass was the strongest single predictor, and the equation doesn’t ask for it.

“BMR” is also a loose label. Basal metabolic rate is strictly measured after sleep, fasted, fully at rest and at a comfortable temperature; resting metabolic rate (RMR) is measured under looser conditions and runs about 10% higher. Mifflin–St Jeor was fitted to resting measurements, so a “BMR” from it is really a resting-rate estimate.

One person, three numbers: a worked example

Example inputs (hypothetical): a woman aged 35, 165 cm tall, weighing 68 kg, with a desk job and exercise three times a week.

BMI is weight in kilograms divided by height in meters, squared:

$$\text{BMI} = \frac{68}{1.65^2} = 24.977\ldots$$

At one decimal place, that shows as 25.0. CDC’s healthy-weight band is 18.5 to less than 25, so the unrounded value is inside it, while the display looks like the first value of the overweight band. She would reach exactly 25 at 68.0625 kg, which is 62.5 g more, about a quarter cup of water.

BMR. The same inputs give different answers depending on the equation (rounded to whole numbers):

The female forms, where $W$ is weight in kg, $H$ is height in cm and $A$ is age in years:

• Mifflin–St Jeor: $10W + 6.25H - 5A - 161$
• Revised Harris–Benedict: $447.593 + 9.247W$ ${} + 3.098H - 4.330A$
• Katch–McArdle: $370 + 21.6 \times \text{lean mass}$, with lean mass in kg

For Mifflin–St Jeor: 10 × 68 + 6.25 × 165 − 5 × 35 − 161 = 680 + 1,031.25 − 175 − 161 = 1,375.25. Revised Harris–Benedict gives 1,436.009, which is 60.76 kcal/day (4.4%) higher. Mifflin’s paper also lists unrounded coefficients (9.99 for weight, 4.92 for age); they add just 2 kcal/day.

Part of the reason they disagree is that each was fitted to different people. Harris–Benedict was derived in 1919 from 239 people, and Roza and Shizgal refitted it in 1984 after adding 98 more of Benedict’s subjects; Mifflin–St Jeor comes from 498 adults, 234 of them obese. In Mifflin’s sample, the original 1919 equations overestimated measured resting expenditure by 5%.

Katch–McArdle is only as good as the body-fat reading behind it. Each percentage point moves it by 21.6 × 0.68 = 14.69 kcal/day for this person, so a true figure of 25% or 35% instead of 30% gives 1,472 or 1,325: about 73 kcal/day either way, more than the gap between the other two equations.

TDEE. Multiply the Mifflin–St Jeor result, 1,375.25, by the five conventional activity multipliers the TDEE calculator offers:

Each step is 0.175 × 1,375.25 = 240.67 kcal/day. She picks 1.55 and gets 2,132 kcal/day. A ±10% band around that is 1,918 to 2,345 kcal/day.

Which choice moves the estimate the most?

The activity level, by a wide margin. The table changes one input or assumption at a time in the same example; the TDEE column uses the 1.55 multiplier.

Switching equations moves TDEE by less than half of one activity step, and no choice of equation removes the person-to-person error.

How far off can a BMR equation be for one person?

Often by more than 10%, and not always in the same direction:

• Systematic review (2005): Frankenfield and colleagues found Mifflin–St Jeor landed within 10% of measured resting expenditure for more people than three other common equations, with the narrowest error range, but still with notable individual errors. Older adults and U.S. ethnic minorities were underrepresented in both building and testing the equations.
• Harris–Benedict (1984): the re-evaluation that produced the revised version put the equations’ precision at about ±14% for a normally nourished person.
• 620 healthy women in Chile (2021): Mifflin–St Jeor overestimated measured resting expenditure by about 208 kcal/day on average and was within ±10% for only 205 of them (33%).
• 127 women in a 2023 test group: the same equation underestimated by about 73 kcal/day on average. It missed by 170 kcal/day on average in one direction or the other, and its largest misses were about 26% low and 27% high.

For the example woman, ±10% of 1,375 is 1,238 to 1,513 kcal/day. In the Chilean study, the equation missed by more than that margin for two women in three. Only a measurement, such as indirect calorimetry in a clinic, removes the equation error.

Which activity multiplier should you pick?

The one that matches your whole day, not just your workouts. The multiplier is the same idea as the physical activity level (PAL) used by the FAO/WHO/UNU expert consultation: total daily energy as a multiple of BMR. That report’s bands start well above the conventional 1.2:

The report’s sample sedentary day comes to 1.53: 8 hours asleep, 8 hours sitting at work, an hour each of driving and walking, some chores and 2 hours of TV or conversation. The NIH Body Weight Planner gives typical levels from 1.4 (sedentary) to 2.5 (very active), with a default of 1.6 for mostly seated work plus moderate activity, such as walking or cycling, at least once a week.

The conventional scale calls 1.2 “sedentary”. For the example woman, 1.2 gives 1,650 kcal/day and 1.53 gives 2,104: 454 kcal/day apart for the same desk-job day. The scales aren’t strictly interchangeable, because PAL multiplies a strict BMR and Mifflin–St Jeor predicts the resting rate, about 10% higher. That doesn’t close the gap for her: the report’s own BMR equation for women aged 30 to 60 (8.126 × weight in kg + 845.6) gives 1,398 kcal/day, slightly more than Mifflin–St Jeor’s 1,375, so its sedentary example works out to about 2,139 kcal/day, still far above 1,650.

• Don’t count exercise twice. The multiplier already includes your workouts; adding a fitness tracker’s “active calories” on top counts them again.
• Choose by your typical week, not your most active one, and count your job as well as your workouts. Using the energy costs in the report’s tables, swapping 8 hours of sitting work for 8 hours on your feet carrying light loads raises PAL by about 0.23. Swapping an hour of TV for an hour of low-intensity aerobic exercise raises it by about 0.12. For the example woman, that’s about 321 vs. 160 kcal/day.

What can’t BMI tell you?

What your weight is made of. CDC describes BMI as a screening measure: it doesn’t measure body fat directly, can’t distinguish fat from muscle and bone, and says nothing about where fat is carried. A DEXA scan measures body composition precisely, but CDC notes such scans are expensive and not widely available.

Two adults who are both 180 cm and 90 kg have a BMI of 27.8 whether the extra weight is muscle or fat, and CDC notes that very muscular people can land in the overweight or obesity range. It recommends reading BMI alongside other information, such as blood pressure, cholesterol, laboratory results and a physical exam.

The cut-offs are sharp; the measurements aren’t. For the example woman, a height recorded 2 cm short (163 cm) gives a BMI of 25.6, and 2 cm tall (167 cm) gives 24.4. Within a few tenths of a boundary, a change of category can come from measurement error alone.

Who shouldn’t use the adult BMI categories?

Children, teens and people who are pregnant; the categories can also understate risk in many Asian populations.

• Children and teens (2 to 19). They are still growing, so CDC compares their BMI with others of the same sex and age and uses percentiles instead. For example, “healthy weight” is the 5th to less than the 85th percentile.
• Pregnancy. Weight gain is expected, and CDC’s weight-gain guidance is based on BMI before pregnancy. Energy needs also change by trimester, which the BMR and TDEE equations don’t model.
• Asian populations. A WHO expert consultation found that a substantial share of Asian people have a high risk of type 2 diabetes and cardiovascular disease at BMIs below 25. It kept the international cut-offs but added public health action points at 23, 27.5, 32.5 and 37.5 for countries to use. The example woman’s 24.98 is under 25 but above 23.

For adults 20 and older, CDC uses the same categories regardless of age, sex or race. BMR equations have a similar limit: Mifflin–St Jeor’s sample covered ages 19 to 78, so outside that range the result is an extrapolation.

How should you use these estimates?

As a first guess that your own records then correct:

1. Start with a range. The example woman’s 2,132 kcal/day becomes a starting band of about 1,900 to 2,350 (±10%). Her real maintenance can still fall outside it, which the next two steps check.
2. Observe for several weeks. Track average daily intake and a weekly average of your weight, which smooths out day-to-day swings.
3. Let the record win. If her weight holds steady for four weeks while she averages, say, 2,000 kcal/day (example figure), about 2,000 is her observed maintenance, 132 kcal/day below the estimate and inside its band. The NIH Body Weight Planner notes that people often underestimate how much they eat, so a food log can read low.
4. Re-check after your weight changes. Mifflin–St Jeor drops 10 kcal/day for each kg lost, and the body adapts too: Hall and colleagues found the weight response to a lasting change in intake plays out with a half-time of about a year. The NIH Body Weight Planner models that adaptation; a fixed multiplier can’t.

These are population estimates, not medical advice. If you’re pregnant, under 20, managing a health condition or unsure what a number means for you, a clinician or registered dietitian can read it alongside the rest of your health information. The methodology page explains how the site checks its formulas.

Try it

• Adult BMI Calculator: choose metric units and enter weight 68 kg, height 165 cm and age 35. The BMI shows as 25.0, a note gives the unrounded value as 24.98, and the category is healthy weight because the calculator classifies the unrounded value. Change the height to 167 cm to see 24.4, then to 163 cm to see 25.6.
• BMR Calculator: choose metric units, set sex used by the equation to female and enter age 35, weight 68 kg and height 165 cm. Mifflin–St Jeor gives about 1,375 kcal/day and revised Harris–Benedict about 1,436. Enter a body fat of 30% for Katch–McArdle (about 1,398), then try 25% (1,472) and 35% (1,325).
• TDEE Calculator: choose metric units and enter the same four inputs, or set BMR to a known value of 1,375.25 kcal/day. Choose moderately active (× 1.55) to get about 2,132 kcal/day. Step through the other levels, then choose a custom multiplier of 1.53 to see about 2,104.