Mifflin vs Harris-Benedict: Which BMR Formula Should You Actually Use?

Where the Harris-Benedict Equation Came From, and Why That Matters

James Arthur Harris and Francis Gano Benedict published their equation for estimating basal metabolism in 1919, and it was a genuine scientific achievement for its era. ^[1] They collected respiration data from hundreds of subjects at the Carnegie Institution Nutrition Laboratory and fit separate regression equations for men and women using age, weight, and height. For the early twentieth century, this was rigorous work.

The problem is the sample. The subjects were predominantly young, healthy, white American adults, a population whose body composition, activity patterns, and average weight differ substantially from what a general-population formula needs to handle today. Equations derived from narrow samples produce predictions that drift when applied to populations outside that sample. Decades of clinical use confirmed exactly that.

Roza and Shizgal addressed this directly in 1984, reanalyzing the original data and revising the regression coefficients to correct for known systematic errors. ^[2] The result is what most calculators now label simply as "Harris-Benedict," though it is technically a revision of the original rather than the original itself. That distinction matters enormously, because the two versions produce different outputs, and most users have no idea which one a given calculator is running.

The Mifflin-St Jeor Equation Was Built to Fix a Specific Problem

By the late 1980s, dietitians working in clinical settings had documented a consistent pattern: the Harris-Benedict equation was overestimating resting energy expenditure in their patients, leading to calorie prescriptions that did not match actual metabolic need. The equation was not just imprecise; it was imprecise in a predictable direction.

Mifflin and colleagues published their response in 1990. They recruited 498 subjects across a wide range of ages, body weights, and body compositions, deliberately casting a broader net than the original Harris-Benedict dataset. Resting energy expenditure was measured by indirect calorimetry, the gold standard for this type of assessment, and a new predictive equation was derived from the results. ^[3]

The resulting formula uses the same four inputs as Harris-Benedict: weight in kilograms, height in centimeters, age in years, and sex. The coefficients are different, recalibrated to a more representative modern sample. Take a 35-year-old woman weighing 70 kg at 165 cm. The Mifflin equation estimates her RMR at about 1,395 kcal. The revised Harris-Benedict puts her near 1,455 kcal, and the original 1919 equation lands higher still at roughly 1,466 kcal. That gap of 60 to 70 kcal at the BMR level widens once an activity multiplier is applied to estimate total daily energy expenditure.

The Mifflin equation was not built to be theoretically elegant. It was built because the existing tool was producing clinical errors, and a better-calibrated tool was needed. That pragmatic origin is part of why its accuracy data holds up.

The Accuracy Gap Is Real and Quantified

Frankenfield, Roth-Yousey, and Compher conducted a systematic review in 2005 that compared predictive equations for resting metabolic rate against values measured by indirect calorimetry in healthy adults. The findings were direct: Mifflin-St Jeor was the most accurate equation for healthy nonobese adults, with approximately 82% of its estimates falling within 10% of measured RMR. ^[4] The original Harris-Benedict equation systematically overestimated RMR across the reviewed studies. The revised version performed better than the original but still fell short of Mifflin in the nonobese population.

Ten percent sounds like a small margin. For someone with a true RMR of 1,500 kcal, a 10% error means estimates ranging from 1,350 to 1,650 kcal before any activity multiplier is applied. That 300 kcal range at the RMR level translates to a much wider range at the TDEE level once a sedentary or moderately active multiplier is factored in.

The obese population is where the picture gets more complicated. No single predictive equation performs consistently well across the full range of body sizes reviewed by Frankenfield et al. This is partly a body composition problem: at higher body weights, the ratio of lean mass to fat mass varies considerably between individuals, and all weight-based formulas struggle with that variance. For individuals with obesity, validated equations like Mifflin-St Jeor are still among the better options available without direct measurement, but the confidence interval on any given estimate is wider.

The practical implication is that Mifflin's accuracy advantage over the original Harris-Benedict is real and replicated, not a statistical artifact of one study.

What a 100-Calorie Daily Error Actually Does to Your Results

Abstract accuracy statistics are easier to dismiss than concrete outcomes. So consider what a systematic 100 kcal/day overestimate in the BMR formula actually produces in practice.

If a calculator overestimates your RMR by 100 kcal and you apply a sedentary activity multiplier of 1.2, the TDEE overestimate becomes 120 kcal per day. You set what you believe is a 400 kcal daily deficit. The actual deficit is closer to 280 kcal. Over ten weeks, the difference between those two deficit sizes accumulates to roughly 8,400 kcal. At the commonly modeled rate where approximately 7,700 kcal corresponds to one kilogram of body mass change, that error represents more than one kilogram of expected fat loss that does not materialize. ^[5]

This is why people eating carefully and tracking diligently still stall. The formula error is not dramatic enough to be obvious on any given day, but it is consistent enough to meaningfully undercut results over a multi-week diet. Switching from the original Harris-Benedict to Mifflin-St Jeor does not guarantee results, but it removes one identifiable source of systematic error from the chain.

Muscle Mass Changes the Math in Ways Neither Equation Fully Captures

Both Mifflin-St Jeor and Harris-Benedict share a structural limitation: they use total body weight, not lean body mass. This matters because skeletal muscle is the primary metabolic engine at rest. Research by Zurlo and colleagues demonstrated that skeletal muscle metabolism accounts for a disproportionate share of resting energy expenditure relative to its mass. ^[6] Fat tissue, by contrast, is metabolically quiet.

The consequence is predictable. Two people with identical age, height, weight, and sex can have meaningfully different resting metabolic rates if their body compositions diverge substantially. A 190-pound man carrying 15% body fat has more metabolically active tissue than a 190-pound man carrying 30% body fat. Both men receive the same Mifflin-St Jeor estimate. One of those estimates is reasonably accurate; the other is probably too high.

For most of the general population, sitting somewhere near average body composition, this limitation does not produce a dramatic error. The equations were calibrated on a population that spans a realistic range, and they perform reasonably across that range. The problem surfaces at the extremes: competitive athletes carrying unusual amounts of lean mass, and individuals with obesity where fat mass comprises a large share of total weight.

For athletes or anyone with a reliable body fat measurement, the Katch-McArdle formula offers a different approach. It uses fat-free mass directly as its input, bypassing the total-weight problem entirely. In populations where body composition is accurately known, Katch-McArdle may outperform both weight-based equations. The catch is that accurate body fat measurement is harder than it sounds; consumer methods like bioelectrical impedance carry their own error margins, and using an imprecise fat percentage to calculate fat-free mass can produce an estimate no better than Mifflin.

The Original Harris-Benedict Should Be Retired

The case against the original 1919 Harris-Benedict equation is not that it is old. Age alone does not disqualify a formula. The case is that it is less accurate than its own revised version from 1984 ^[2] and less accurate than the Mifflin equation that replaced it in clinical dietetics. ^[4] There is no population subgroup for which the original outperforms either the Roza-Shizgal revision or Mifflin-St Jeor.

Despite this, the original Harris-Benedict coefficients still appear in calculators across the web, sometimes labeled correctly, sometimes not. A calculator that uses the original without disclosing which version it is running is delivering a result that was superseded twice over and that systematically overestimates calorie needs. Users deserve to know which formula a tool is using. If a calculator does not disclose this, that is worth treating as a red flag about its overall rigor.

Where Harris-Benedict (Revised) Still Has a Legitimate Place

The argument above should not collapse into a claim that the revised Harris-Benedict is worthless. It is not. The accuracy gap between the Roza-Shizgal revision and Mifflin-St Jeor at the BMR level is often 50 to 80 kcal, which is smaller than the error introduced by any activity multiplier. Someone estimating their TDEE using a "lightly active" multiplier is already working with an approximation that could be off by several hundred calories depending on how accurately that descriptor maps to their actual daily movement.

The revised Harris-Benedict also remains embedded in clinical nutrition software and hospital dietetics protocols that have not been updated to reflect the 2005 Frankenfield findings. Switching institutional tools requires regulatory clearance, budget, and retraining, none of which happens quickly. For clinicians working within those constraints, the revised Harris-Benedict is not a failure of rigor; it is a pragmatic reality.

For any individual who will track their intake and adjust based on what their body weight actually does over several weeks, the formula choice is secondary. The person who picks Mifflin, sets a target, and never revisits it will get worse results than the person who picks the revised Harris-Benedict, sets a target, and adjusts when the number is not working.

How Metabolic Adaptation Limits What Any Formula Can Predict

Both equations share a deeper limitation that accuracy comparisons do not capture: they estimate RMR at a single point in time, under conditions of energy balance. Sustained caloric restriction changes that baseline.

Adaptive thermogenesis refers to the downward shift in resting energy expenditure that occurs during prolonged dieting, beyond what would be predicted by the loss of lean and fat tissue alone. Rosenbaum and Leibel documented this phenomenon and its persistence, showing that the body actively resists energy deficit through mechanisms that go beyond simple mass reduction. ^[7] This is not a minor effect. Research following contestants from The Biggest Loser found that metabolic adaptation persisted six years after the competition ended, with resting metabolic rates remaining suppressed far below what body composition alone would predict. ^[8]

What this means practically: a Mifflin-based TDEE calculated before a diet begins becomes progressively less accurate as the diet continues. The equation does not know you have been in a deficit for eight weeks. It does not adjust for the fact that your body has downregulated thermogenesis in response. By week twelve of an aggressive cut, the gap between your estimated and actual calorie needs may be driven more by adaptation than by any difference between Mifflin and Harris-Benedict.

This is not an argument against using these equations. It is an argument for treating their outputs as estimates that require periodic recalibration against observed results, not as fixed prescriptions.

The Recommendation Is Simple: Use Mifflin, Then Measure

Mifflin-St Jeor is the right default for healthy adults estimating TDEE for fat loss or muscle gain. The accuracy data supports it, the revised Harris-Benedict is a defensible but inferior second choice, and the original should be treated as obsolete. ^[4]

The BMR Calculator on this site uses Mifflin-St Jeor as its primary formula, with the revised Harris-Benedict available as an alternative for comparison. Running both numbers takes thirty seconds and immediately shows how much the formula choice shifts the output for a given set of inputs.

Once a target is set, four to six weeks of consistent intake and daily weigh-ins will reveal whether the estimate is accurate. Average the daily weights by week to smooth noise. If the trend does not match the expected rate of change, adjust the calorie target by 100 to 150 kcal and run another observation period. The formula provides a calibrated starting number. The observation period tells you whether that number fits your actual physiology.