No single BMR formula is most accurate for everyone. The Mifflin-St Jeor equation is the most widely recommended starting point, but its accuracy depends heavily on your age, body composition, and weight. In studies, even the best formulas only predict within 10% of your actual metabolic rate about 50% to 60% of the time, which means any calculator you find online could be off by a few hundred calories in either direction.
The real answer is that different formulas win for different populations. Here’s what the research actually shows.
Why Mifflin-St Jeor Gets Recommended Most
The American Dietetic Association reviewed the major BMR prediction equations and concluded that Mifflin-St Jeor was the most likely to land within 10% of a person’s actual resting metabolic rate. That recommendation became the default in nutrition guidelines, and it’s why most online calculators use this formula. It factors in your weight, height, age, and sex.
In a recent study of young adult women across different weight categories, Mifflin-St Jeor achieved 60.1% accuracy (meaning 60% of predictions fell within 10% of the measured value). Its average bias was only about 2.9%, and it showed the lowest mean difference from measured values across weight categories, ranging from roughly 16 calories under in underweight individuals to about 84 calories over in those with obesity. For the general adult population under 65, it’s a solid default choice.
Harris-Benedict Isn’t Far Behind
The original Harris-Benedict equation, published in 1919, is the oldest formula still in wide use. Despite its age, it performs surprisingly well. One large study that directly compared the major equations found that Harris-Benedict was actually the most accurate overall, predicting 57.5% of participants within 10% of their measured metabolic rate, compared to 56.4% for Mifflin-St Jeor. The difference is slim enough that neither formula has a meaningful edge in the general population.
A 1984 revision by Roza and Shizgal re-examined the original data with additional subjects spanning a wider age range. They found the revised equations produced similar results to the originals, with a precision of about 14%. So if a calculator uses the “revised Harris-Benedict,” you’re getting a marginally updated version of essentially the same formula.
Where Harris-Benedict falls short is in older adults. In people aged 65 and up, it was the worst-performing predictor among commonly used equations, producing the largest errors. This makes sense: very few older adults were included in the data used to build it over a century ago.
Which Formula Works Best for Obesity
If your BMI is over 30, the picture shifts. In a study of older adults with obesity, the WHO equation had the highest accuracy at 59%, followed by Harris-Benedict at 53.5%. Mifflin-St Jeor dropped to just 43.1% accuracy in that same group. The Owen equation, while only 50.7% accurate, had the most balanced error profile with roughly equal rates of over- and underprediction.
That said, in a younger cohort of women with obesity, Mifflin-St Jeor performed much better, reaching 83.3% accuracy. The takeaway: age and obesity together make prediction harder. A formula that works well for a 30-year-old with obesity may struggle for a 70-year-old with the same BMI. All standard formulas use total body weight, which doesn’t distinguish between muscle and fat. Since fat tissue burns fewer calories than muscle, two people at the same weight can have very different metabolic rates.
For Athletes and Muscular Builds
If you carry significantly more muscle than average, formulas based on lean body mass should theoretically outperform weight-based ones. The Cunningham equation and Katch-McArdle equation both use lean body mass (your weight minus your fat mass) as the primary input instead of total weight.
A 2023 systematic review of metabolic rate equations in athletes found something unexpected. The Cunningham equation predicted 54.1% of athletes within 10% of measured values, while the standard Harris-Benedict hit 53.7% and Mifflin-St Jeor landed at 52.2%. None of these were significantly different from each other, and all showed enormous variability, with the same formula overpredicting for some athletes and underpredicting for others.
The standout performer in athletes was actually the Ten-Haaf equation from 2014, which doesn’t require lean body mass at all. It uses age, weight, and height, yet predicted 80.2% of athletes within 10% of their measured metabolic rate. Every other equation ranged from 40.7% to 63.7%. It also showed zero heterogeneity across studies, meaning it performed consistently rather than swinging wildly between over- and underprediction. If you’re an athlete or highly active person, this lesser-known formula may be worth seeking out.
Accuracy Drops After Age 65
Standard BMR equations were built primarily on data from younger adults. The Mifflin-St Jeor equation, for example, included only 15 men aged 65 to 79 and no one over 80. As you age, your body composition changes: muscle mass declines, organ mass shifts, and metabolic rate drops in ways the standard formulas don’t fully capture.
Researchers recently developed new equations specifically for adults 65 and older. These equations, using the same simple inputs of weight, height, and age, achieved a population-level prediction bias of only about 12 calories per day (roughly 1% error) for those under 80, rising to about 24 calories per day (2% error) for those 80 and above. For men over 80, weight alone was enough to predict metabolic rate reasonably well, and sex was no longer a significant factor in the model.
Even these improved equations had wide individual-level margins of error, with 95% confidence limits stretching to roughly plus or minus 25%. Population averages look great, but for any single person, the prediction could still be substantially off.
Every Formula Has the Same Core Problem
The gold standard for measuring metabolic rate is indirect calorimetry, a clinical test where you breathe into a device that measures your oxygen consumption and carbon dioxide output. Every prediction equation is benchmarked against this method, and none of them consistently match it at the individual level.
In underweight women, most common equations overestimated metabolic rate by 42 to 224 calories per day compared to indirect calorimetry. Only specialized equations like the Muller formula came close, with an average difference of just 2.8 calories per day. The pattern repeats across populations: formulas work reasonably well on average but can miss by 200 or more calories for any given person.
This matters practically. If you’re using a BMR calculation to set calorie targets and the formula is off by 150 to 200 calories, that error compounds over weeks. It’s why nutrition professionals treat these numbers as starting estimates, not precise measurements. Your actual metabolic rate is influenced by genetics, thyroid function, sleep quality, stress hormones, and dozens of other variables that no simple equation can account for.
Choosing the Right Formula for You
- General adults under 65: Mifflin-St Jeor or Harris-Benedict. Both predict within 10% accuracy for roughly 55% to 60% of people. Mifflin-St Jeor has the broader clinical endorsement.
- Athletes or highly active people: The Ten-Haaf equation outperforms all others at 80.2% accuracy. The Cunningham equation is a reasonable alternative if you know your lean body mass, but it doesn’t actually outperform simpler formulas in most studies.
- Adults with obesity: Results vary by age. For younger adults, Mifflin-St Jeor holds up well. For older adults with obesity, the WHO equation tends to perform better.
- Adults over 65: Standard equations lose accuracy. Newer age-specific equations exist but aren’t yet built into most online calculators. Harris-Benedict performs the worst in this age group.
Whatever formula you use, treat the result as an estimate with a margin of error of at least 10% in either direction. If you calculate a BMR of 1,500 calories, your actual value likely falls somewhere between 1,350 and 1,650. Use the number as a starting point, then adjust based on what actually happens with your weight and energy levels over two to four weeks.