Mifflin-St Jeor is the BMR equation with the best validation record, which is why our calculator uses it — but "best" here means "wrong by the smallest amount for the most people", not "right for you". Harris-Benedict runs about 5% high against modern measurements, and the lean-mass equations (Cunningham, Katch-McArdle) are only as good as the body-fat number you feed them. Here is how the four compare, with the receipts.
The four equations side by side
| Equation | Year · sample | Inputs | Formula (kcal/day) |
|---|---|---|---|
| Mifflin-St Jeor1 | 1990 · 498 adults, 19–78 y | weight, height, age, sex | men 10w + 6.25h − 5a + 5 · women 10w + 6.25h − 5a − 161 |
| Harris-Benedict2 | 1918 · 136 men, 103 women | weight, height, age, sex | men 66.4730 + 13.7516w + 5.0033h − 6.7550a · women 655.0955 + 9.5634w + 1.8496h − 4.6756a |
| Cunningham3 | 1980 · 223 subjects (re-analysis of the Harris-Benedict data) | lean body mass | 500 + 22 × LBM |
| Katch-McArdle | textbook equation (McArdle, Katch & Katch, *Exercise Physiology*) | lean body mass | same family as Cunningham — a constant plus a coefficient on lean mass; we have not verified the printed coefficients against the book, so they are not reproduced here |
w = weight in kg, h = height in cm, a = age in years, LBM = lean body mass in kg.
Two families, then. Mifflin-St Jeor and Harris-Benedict describe you by size, age and sex. Cunningham and Katch-McArdle describe you by how much non-fat tissue you carry, which is the tissue that actually spends most of the resting energy — Cunningham's re-analysis found lean body mass to be the single predictor of BMR, with sex and age adding little once it was known3.
What each one needs as input
- Weight, height, age, sex — everyone has these. They are what Mifflin-St Jeor and Harris-Benedict need, and they are why those two are the default in almost every calculator, including ours.
- Lean body mass — you need a body-fat percentage to get it:
LBM = weight × (1 − body fat). A skinfold, a DEXA scan or a bioimpedance scale gives a number; a guess gives a guess.
The trade-off is exact. The lean-mass equations skip the population averages for sex and age, but they import whatever error is in your body-fat figure. Five percentage points of body-fat error on an 85 kg person is 4.25 kg of lean mass, which in Cunningham's equation is 22 × 4.25 ≈ 94 kcal of BMR. Most people's body-fat estimate is off by more than five points.
How wrong they typically are
For our standard example — man, 30 years, 180 cm, 85 kg:
| Equation | BMR (kcal/day) | Note |
|---|---|---|
| Mifflin-St Jeor | 1,830 | our calculator's value |
| Harris-Benedict | 1,933 | 66.4730 + 13.7516×85 + 5.0033×180 − 6.7550×30 |
| Cunningham at 15% body fat | 2,090 | LBM = 72.25 → 500 + 22 × 72.25 |
| Cunningham at 25% body fat | 1,903 | LBM = 63.75 → 500 + 22 × 63.75 |
Harris-Benedict lands 103 kcal above Mifflin-St Jeor here — 5.6% — which matches what Mifflin and colleagues found across their whole sample: the 1918 equations overestimated measured resting expenditure by 5%1. Cunningham swings by 187 kcal between two plausible body-fat guesses for the same man, which is the lean-mass trade-off in one row.
Against measurement rather than against each other, the picture is humbler still. Mifflin-St Jeor explains 71% of the variance in measured resting expenditure in its own development sample1. The 2005 systematic review by Frankenfield and colleagues, which pooled validation studies reporting individual errors, concluded that Mifflin-St Jeor was the most reliable of the four equations in common clinical use — predicting resting metabolic rate within 10% of measured in more non-obese and obese adults than Harris-Benedict, Owen or WHO/FAO/UNU, with the narrowest error range — and in the same breath that noteworthy errors and limitations exist when it is applied to individuals4.
Put a number on "within 10%": for a 1,830 kcal BMR that is a ±183 kcal band, and "most people" inside the band means a meaningful minority outside it. As a working rule, a population equation is typically 100–200 kcal off for an individual, and 400–500 kcal misses are not rare. This is why we describe every calculator output as a starting estimate and why the app's expenditure model replaces it with your own data as soon as there is enough of it.
In plain terms: the equations differ from each other by about a hundred kcal; each of them can differ from you by several hundred.
Why we use Mifflin-St Jeor
Three reasons, in order:
- It has the best validation record. The Frankenfield review is the most systematic head-to-head we have, and it favours Mifflin-St Jeor on both the share of people within 10% and the width of the error range4.
- It needs only what everyone knows. Sex, age, height, weight. A calculator that asks for body fat gets a guess back from most users, and a lean-mass equation fed a guess is not more accurate — it is differently wrong.
- It was fitted on people like the people who use calculators. Mifflin's sample was half women and half people with obesity, aged 19–781; the Harris-Benedict sample was 239 mostly young, mostly lean adults measured a century ago2.
None of that makes it correct for you. It makes it the least-bad first guess, which is all a first guess has to be.
This is a comparison of population formulas, not medical advice. The differences between equations are smaller than the difference between any equation and a given person — which is why two weeks of logged intake and trend weight beats all four.
The short version
Four equations, two families: Mifflin-St Jeor and Harris-Benedict from size, age and sex; Cunningham and Katch-McArdle from lean mass. For our 85 kg example they land between 1,830 and 2,090 kcal. Mifflin-St Jeor has the best validation record and needs no body-fat guess, so it is our default — and it is still typically 100–200 kcal off for any one person. Educational overview only — not medical advice.