The number your calorie calculator gives you runs on an equation derived from 498 adults. It replaced math built on a 1919 population averaging 64.1 kg. It explains 71% of your metabolism. The other 29% is invisible to any equation ever published.
Nearly a third of what actually determines your resting metabolism is invisible to the math producing that number.
You have probably never questioned the number your calorie calculator gives you. You typed in your weight, height, age, and sex. The app ran the math. A number appeared on screen, and you started building your meals around it.
When the scale didn't move after three weeks, you blamed yourself. Your snacking. Your consistency. Your willpower at 9 PM. Across fitness communities on Reddit, the same frustration plays out every week: someone tracks every bite, hits their calorie number, and the result doesn't match the math. The advice is always the same. Track more carefully. Weigh your food. Be more disciplined.
Almost nobody asks the upstream question: how accurate was that number in the first place?
That number runs on an equation. And the equation most likely to be running behind your calculator has a story that starts more than a century ago.
In 1919, researchers at the Carnegie Nutrition Laboratory published a set of formulas for predicting how many calories a body burns at rest. The Harris-Benedict equations were derived from measurements of 136 men and 103 women. Those people averaged 64.1 kilograms (141 lbs).
The study population measured seven decades later averaged 87.5 kilograms (193 lbs). That is a 36% shift in body size, and the equation had never been recalibrated for it.
A note of precision: the original 1919 coefficients were revised in 1984 by Roza and Shizgal, and most modern calculators use that updated version. The math got sharper. But the calibration population (the bodies the equation was trained on) never changed. The coefficients were refined. The foundation was still 1919.
A team led by Mark Mifflin and Sachiko St Jeor tested the Harris-Benedict equations against actual measurements from 498 adults. The result was clear: the old equations overestimated how many calories a body burns at rest by 5%.
For someone eating to a 2,000-calorie number based on that math, that is roughly 100 phantom calories baked into the baseline. Over a year, that gap adds up.
The best calorie equation science has found predicts within 10% of your actual resting metabolism for most people, but the equation fitness communities call "more advanced" turned out to be the least accurate of all four tested.
- A team measured resting metabolism in 498 adults and built the equation most calorie apps now use. It captures 71% of the variation in how many calories a body burns at rest.
- A 2005 systematic review found the equation predicts within 10% of the measured value for 82% of non-obese people. For obese individuals, that accuracy drops to 70%.
- The Cunningham equation, often presented as the smarter option because it uses lean body mass, overestimated resting calorie burn by 14 to 15% in this study. The "advanced" choice ranked last.
- About 29% of what determines your resting metabolism is invisible to any equation the researchers could build. That gap is why the number on your calculator is a starting point that works best when paired with what your body actually shows you over time.
The Team That Rebuilt the Math
The study that produced the replacement was published in 1990 in the American Journal of Clinical Nutrition. Mifflin and St Jeor recruited 498 healthy adults in the United States (247 women, 251 men) ranging in age from 19 to 78, with BMIs from 17 to 42. Both normal-weight and obese individuals were included by design.
Every participant's resting energy expenditure (how many calories the body burns at rest, before exercise) was measured directly using indirect calorimetry. A hood placed over the participant's head captured oxygen consumption and carbon dioxide output while they rested. This was not a questionnaire or an estimate. It was a measurement.
From those 498 measurements, the researchers derived a new equation: 10 times your weight in kilograms, plus 6.25 times your height in centimeters, minus 5 times your age, plus 5 for men or minus 161 for women. A simplified version with rounded numbers performed identically (the correlation between the two was 1.0).
If you use MyFitnessPal, you are probably already running on this equation. The app adopted Mifflin-St Jeor as its default, joining a shift that has reached more than 220 million downloads worldwide. A 2005 systematic review confirmed it as the most reliable of the commonly used equations [1].
So the better equation won. Your app is almost certainly using it. That should be the end of the story.
It is not.
The 29% No Equation Can Capture
The Mifflin-St Jeor equation explains 71% of the variation in resting energy expenditure across those 498 people. That means the four variables it uses (weight, height, age, sex) account for roughly seven out of every ten units of metabolic variation in the sample.
The other 29% is a genuine scientific mystery.
The researchers tried to close that gap. They added body fat percentage. They added body mass index. They added waist-to-hip ratio. They built separate equations for men and women. They tested every combination they had. None of it improved the prediction beyond 0.71.
The combined equation using four simple variables outperformed every specialized version. Separate equations for men alone explained 47% of the variation. For women alone, 62%. Both worse than the combined formula's 71%.
What accounts for the missing 29%? The study's authors pointed to genetic differences in metabolic efficiency and acquired variations from factors like chronic dieting history. That is where the data stops. The variables they measured, and they measured everything practical, could not explain it.
If your calorie calculator gives you a number, nearly a third of what actually determines your resting metabolism is invisible to the math producing that number. Not because the equation is bad. Because the phenomenon is that complex.
One of those invisible components: involuntary movement that varies tenfold between individuals — when 16 people ate the same extra 1,000 calories per day, their fidgeting, not their resting metabolic rate, predicted nearly everything about who gained fat.
The equation fitness communities call more advanced overestimated resting energy expenditure by 14 to 15 percent. More data did not mean more accuracy.
The "Advanced" Equation That Performed Worst
After learning that the best equation has a 29% blind spot, the natural reaction is to look for a smarter one. In fitness communities, the Cunningham equation is often presented as the upgrade. It uses fat-free mass instead of total weight, height, and age. The logic sounds airtight: if muscle burns more calories than fat, an equation that accounts for body composition should be more accurate.
The Mifflin-St Jeor study tested that logic directly. The Cunningham equation overestimated resting energy expenditure by 14 to 15 percent in this population. That made it the worst performer of the four equations tested. Not the best.
An important caveat: this study measured healthy adults across a wide range of body sizes, not competitive athletes. For lean, heavily muscled individuals, the Cunningham equation may perform differently. But for the vast majority of people typing their stats into a calculator, the equation marketed as more sophisticated was the least accurate.
The Owen equations came closest to the actual measured values: within 4% for women and 0.1% for men. Even those did not outperform Mifflin-St Jeor's overall predictive power. The simplest practical formula, using four variables you already know without any lab measurement, remained the best available tool.
More data did not mean more accuracy. The hierarchy fitness communities assume was inverted by the actual measurements.
How Accurate Is the Best Available Number
If the best equation still has a 29% blind spot and the "advanced" alternative performed worst, the question becomes practical: how much can you actually trust the number on your screen?
Fifteen years after the Mifflin-St Jeor equation was published, a systematic review by Frankenfield and colleagues evaluated it against measured resting metabolic rates across multiple independent studies [1]. Their finding: it predicts within 10% of the actual measured value for 82% of non-obese individuals. For obese individuals, that accuracy drops to 70%.
That is both reassurance and a reality check in the same breath.
The reassurance: 82% accuracy within 10% makes Mifflin-St Jeor the most reliable of the commonly used equations. No competing formula matched it across both weight categories. For most people, the number your app gives you is in the right neighborhood.
The reality check: for 18% of non-obese people and 30% of obese people, the equation falls outside that 10% margin. If you are in that group, the number on your screen could be meaningfully wrong, and you would have no way of knowing from the calculator alone.
The equation's creators and the systematic review's authors were both honest about where the math falls short. That honesty is part of the science, not an inconvenient footnote to hide.
The 29% of metabolism that no equation can capture is not a failure. It is a map of where the math stops and your own body's data begins. What sits beyond it (genetic differences in metabolic efficiency, acquired variations from dieting history) may narrow as measurement technology improves. Unexplained does not mean unexplainable.
82% accuracy within 10% makes it the most reliable of the commonly used equations. For the other 18%, the equation falls outside that margin.
What Your Calorie Number Actually Means
The number on your screen is the best available starting point. Not a prescription. Not a verdict. A starting point.
The Mifflin-St Jeor equation replaced math built on bodies from a different century, outperforms every alternative tested against it, and predicts within 10% for a clear majority of people. That is genuinely useful.
What it cannot do is account for the whole picture. The 29% of resting metabolism that no equation captures means your actual calorie needs include a component the calculator cannot see. The practical takeaway: treat the number as a place to start, then adjust based on what your body actually does over two to three weeks.
You now know more about the number on your phone than most people who use it every day. You know where it came from, what it replaced, what it can explain, and where it goes blind. The equation is a tool. A good one. Not a perfect one.
Which raises the next question. If the number your calculator gives you is an approximation, and nearly a third of your metabolism is invisible to the math, are you even tracking what you actually eat correctly?
The equation produces a number. The research positions that number as a starting line, not a destination.
What sits beyond the equation's reach includes factors like sleep patterns, stress, habitual movement, and individual metabolic quirks that vary from person to person. The study's authors pointed to genetic and acquired differences in metabolic efficiency as likely contributors to the 29% no formula could capture.
The practical weight of the finding: a calorie number built from weight, height, age, and sex captures most of the picture. The part it misses shows up when the number meets real life. Two to three weeks of observing how your body responds to a given calorie level reveals whether the equation landed close or needs adjustment. The equation did its job by narrowing the range. Your body finishes the calibration.
What other research found
What this means for you
The accuracy picture shifts when BMI enters the obese range. A systematic review found the equation predicts within 10% of measured resting metabolism for 70% of obese individuals, compared to 82% for non-obese.
That means the equation is still the best available starting point, but the margin of uncertainty is wider. For 30% of obese individuals, the number on the screen falls outside 10% of what the body actually burns at rest.
The study's population included 234 obese adults specifically. The equation was not built for lean people only. But its accuracy is highest for them.
The study measured healthy adults across a wide range of body sizes, not competitive athletes. For someone with significantly more muscle and less body fat than the general population, the equation's assumptions shift.
Fat-free mass was the strongest single predictor of resting metabolism in the data. The equation that uses it (Cunningham) performed worst in the general population, but the researchers noted it was never tested against lean, heavily muscled individuals.
The finding applies to the general population. Whether it holds for someone at the athletic extreme is a question this study specifically could not answer.
The 29% of resting metabolism that no equation can capture includes what the researchers called acquired metabolic efficiency, the body's learned response to prolonged energy restriction.
If you have spent months or years cycling through calorie deficits, your resting metabolism may have adapted in ways the equation cannot detect. The gap between the predicted number and your actual burn could be wider than average.
This study did not measure dieting history directly. But the authors named acquired efficiency as one likely component of the unexplained variance, making it relevant to anyone with an extended cutting history.
Before you change anything
498 healthy adults in the United States, ages 19 to 78, with body weights ranging from lean to moderately obese (BMI 17 to 42). Both men and women were included in roughly equal numbers (247 women, 251 men).
The participants were recruited from a working-class population, where the subject or their spouse was employed at least half-time. Racial and ethnic background was not reported, so the equation's accuracy across different ethnic populations is an open question.
People who were severely underweight or morbidly obese were largely absent from the sample. So were competitive athletes and adults older than 78. If you fall outside the range of generally healthy, ages 19 to 78, BMI roughly 17 to 42, the equation's accuracy for you specifically has not been directly tested in this study.
The equation was built and tested on the same group of people. That means it was never validated against a separate population in this paper. Independent validation came 15 years later in the Frankenfield systematic review, but the original study itself acknowledged the limitation.
Body composition was measured using skinfold calipers, which the researchers themselves noted pose multiple problems, especially for individuals with higher body fat. More precise methods like DEXA exist but were not practical for 498 people in 1990.
The study was conducted in 1990. Average body weights, activity levels, and dietary patterns in the United States have shifted since then. The equation fit the 1990 population better than the 1919 equation did, but the population has continued to change.
Strong enough to be the standard. Multiple independent research teams have tested this equation since 1990. The American Dietetic Association recommended it as the most reliable option based on a systematic review. It is the default in the most widely downloaded calorie tracking app in the world.
Honest enough to acknowledge the gap. Even with that track record, 18% of non-obese predictions and 30% of obese predictions land outside 10% of what the body actually burns at rest. For most people, the number gets you in the right neighborhood. For a meaningful minority, it does not, and there is no way to know which group you fall into from the equation alone.
You know the equation is a starting point now. You might even adjust the number based on what your body does over the next few weeks. But that adjustment depends on one assumption you have not questioned yet: that the food diary you are building around the equation's output is accurate.
A 1992 study used a measurement method that bypasses self-report entirely to test that assumption. The gap it found between what people said they ate and what they actually consumed makes the equation's 29% blind spot look modest by comparison.
What This Study Found
All findings from this paper, in plain language.
- The equation captures 71% of what determines resting calorie burn using just weight, height, age, and sex.
- Fat-free mass was the single strongest predictor of resting metabolism, but measuring it requires equipment most people do not have.
- The older Harris-Benedict equation overestimated resting calorie burn by 5% compared to what the researchers actually measured.
- The Cunningham equation, which uses lean body mass, overestimated by 14 to 15% and was the worst result of any equation tested.
- The Owen equations came closest to measured values after Mifflin-St Jeor, predicting within 4% for women and 0.1% for men.
- Adding body fat percentage, BMI, and waist-to-hip ratio to the equation did not improve its accuracy beyond 71%.
- Building separate equations for men and women did not outperform the combined equation that uses sex as a single variable.
- About 29% of resting metabolic variation remained unexplained by any combination of variables the researchers tested.
- The original Harris-Benedict sample averaged 64 kilograms for men, compared to 87.5 kilograms in this study, a 36% shift in body size.
- Rounding the equation's numbers for easier use did not change its accuracy at all, and the simplified version matched perfectly.
- The equation's sample better reflects modern body weights than the 1919 population the Harris-Benedict equations were built on.