Mifflin-St Jeor vs Katch-McArdle: Which BMR Equation Should You Actually Use?

Every macro calculator starts with a resting metabolic rate estimate, and that estimate depends entirely on which equation runs behind the scenes. Mifflin-St Jeor and Katch-McArdle are the two most commonly debated options, and they can disagree by well over 100 kcal/day for the same person. This article walks through where each formula came from, what the validation data actually shows, and which one you should trust for your own numbers.

If you want the answer applied to your own stats without doing the arithmetic by hand, the Scinergy macro calculator defaults to the formula with the best population-level evidence and lets you switch to a lean-mass-based estimate if you have a reliable body fat measurement.

Where these equations came from: a short history of BMR prediction

Harris-Benedict (1919) is the origin point for modern BMR prediction. J. Arthur Harris and Francis G. Benedict at the Nutrition Laboratory of the Carnegie Institution of Washington measured basal metabolism via indirect calorimetry in 136 men, 103 women, and 94 newborn infants, publishing first in Proceedings of the National Academy of Sciences in 1918 and then the full monograph the following year (PNAS, 1918; PMC record; 1919 monograph, Internet Archive). Their original regression had an R-squared of only 0.64 for men and just 0.36 for women, a notably weak statistical fit even by the standards of the time (Metabolites, 2023). The equations were revised by Roza and Shizgal in 1984 with an expanded dataset, producing the coefficients most commonly seen today and improving the fit to an R-squared of 0.77 for men and 0.68 for women (TDEE Calculator Kit history page).

Mifflin-St Jeor (1990) was developed specifically to correct Harris-Benedict's overestimation in modern populations. Mifflin, St Jeor, and colleagues derived their equation from 498 healthy subjects of both sexes, ages 19 to 78, using indirect calorimetry, publishing in the American Journal of Clinical Nutrition in 1990 (Mifflin et al. 1990, full text; DOI record). Their key finding was that the original 1919 equations overestimated measured resting energy expenditure in this modern sample by 5 percent (p less than 0.01). Their final combined-sex regression achieved an R-squared of 0.71, and simplified by sex it becomes the now-standard form: for men, 10 times weight in kg plus 6.25 times height in cm minus 5 times age plus 5; for women, the same terms minus 161 instead of plus 5. Interestingly, the same 1990 paper found fat-free mass alone was actually the single best predictor of resting energy expenditure in their dataset, with an R-squared of 0.64, which is the direct intellectual ancestor of lean-mass-based equations like Katch-McArdle.

Katch-McArdle traces its roots to lean-body-mass-based prediction models developed in exercise physiology. Frank Katch and William McArdle proposed BMR equal to 370 plus 21.6 times lean body mass in kg in their textbook "Exercise Physiology," with sources dating the underlying formula development to somewhere between the mid-1970s and 1996 depending on edition (CalcoI.com formula history; Fit Life Regime). It is closely related to the Cunningham equation from 1980, BMR equal to 500 plus 22 times lean body mass, itself derived to correct known lean-mass prediction issues in athletic populations (PeakCalcs formula comparison). Both share the same logic as the Mifflin 1990 fat-free-mass regression: because muscle tissue is metabolically active and fat tissue is comparatively inert, a formula built purely on lean mass should, in theory, generalize better across body composition extremes than one built on total body weight.

The accuracy data: what validation studies actually found

The single most important validation study in this space is Frankenfield, Roth-Yousey, and Compher (2005), a systematic review published in the Journal of the American Dietetic Association that analyzed data across roughly 1,090 subjects comparing predicted resting metabolic rate to indirect calorimetry (Frankenfield 2005, cited via OUCI; summary via PeakCalcs). We cover the study in full in the next section, but the headline finding is what shaped clinical practice for the last two decades: no equation before or since has matched Mifflin-St Jeor's combination of simplicity and population-level accuracy.

For lean-mass equations, accuracy data is thinner and more population-dependent. Secondary compilations of validation literature suggest Cunningham accuracy around 76 percent within 10 percent, and Katch-McArdle around 73 percent within 10 percent, in smaller studies, rising toward 80 percent-plus when lean body mass is measured via DEXA in genuinely lean populations (CalcFit compiled comparison table). Other work reports correlation coefficients of 0.85 to 0.92 between Katch-McArdle predictions and indirect calorimetry in athletic populations, though that figure should be treated cautiously since the underlying primary study was not independently verified in our research (Fit Life Regime, citing ResearchGate validation data).

Later independent work confirms the Frankenfield-era conclusion. A study of 149 Emirati female young adults across all BMI categories found Mifflin-St Jeor was the most accurate of nine equations tested, while Harris-Benedict was the most inaccurate (study summarized on OUCI). A separate 2020 study in a Slovenian population found the Owen equation, not covered in this article's main comparison, was actually most comparable to measured resting energy expenditure, illustrating that no single equation dominates every population subgroup (Slovenian Journal of Public Health, 2020). A 2023 systematic review and meta-analysis focused on post-bariatric surgery patients found Mifflin-St Jeor had the lowest bias for predicting resting energy expenditure in that population too, with Harris-Benedict also performing acceptably in that specific context (PubMed, 2023).

Frankenfield 2005: the study that made Mifflin-St Jeor the default

Frankenfield's systematic review found that Mifflin-St Jeor predicted resting metabolic rate within 10 percent of measured values in 82 percent of non-obese adults and 70 percent of obese adults, the highest accuracy of any equation tested (CalcFit summary of Frankenfield 2005). Harris-Benedict, by contrast, achieved only 69 percent accuracy in non-obese subjects and a range of just 38 to 64 percent in obese subjects depending on subgroup, with a consistent overestimation bias of 5 to 15 percent. Based on this review, the Academy of Nutrition and Dietetics, then the American Dietetic Association, formally adopted Mifflin-St Jeor as its recommended default equation for healthy adults. That single systematic review is the reason Mifflin-St Jeor, not Harris-Benedict or Katch-McArdle, is the default formula behind most modern calculators, including the Scinergy macro calculator.

Three forces keep the 1919 equation alive anyway despite being effectively obsolete: it was first, so it is deeply embedded in decades of spreadsheet templates and textbooks; it is simple to hand-calculate and easy to teach, so it persists in certification courses long after better data exists; and most calculator builders copy prior calculators rather than consulting the primary validation literature, a kind of calculator inertia with no equivalent correction mechanism (Harris-Benedict history, TDEE Calculator Kit). The same inertia problem shows up in how activity multipliers get chosen, where a similarly old and under-validated ladder of numbers still runs behind most online calculators.

Katch-McArdle and Cunningham: when lean-mass equations win

For the general population, meaning normal BMI with mixed body composition and no known body fat percentage, Mifflin-St Jeor wins. It requires only weight, height, age, and sex, avoids the error-compounding step of estimating body fat percentage, and the Frankenfield 2005 review remains the best available evidence base.

For lean, muscular individuals and trained athletes with a known body composition, Katch-McArdle or Cunningham can outperform weight-based equations, because two people at the same total body weight but different muscle mass have meaningfully different resting energy needs. The catch is that the equation's accuracy is entirely bottlenecked by the accuracy of the lean body mass input, which is the subject of the next section. For obese individuals, Mifflin-St Jeor still performs best among common formulas at 70 percent within 10 percent per Frankenfield, while Harris-Benedict degrades sharply, dropping as low as 38 percent within 10 percent in some subgroups, and tends to substantially overestimate. If lean mass and protein targets are your main concern rather than just the calorie number, our protein for muscle gain guide covers how lean mass factors into daily protein needs directly.

Worked example: 85kg male at 15 percent body fat

Take an 85kg, 178cm, 32 year old man at 15 percent body fat and run all three formulas side by side. Mifflin-St Jeor: 10 times 85 plus 6.25 times 178 minus 5 times 32 plus 5, which works out to 850 plus 1,112.5 minus 160 plus 5, for a total of 1,807.5 kcal/day.

Harris-Benedict, using the revised 1984 coefficients: 88.362 plus 13.397 times 85 plus 4.799 times 178 minus 5.677 times 32, which works out to 88.362 plus 1,138.75 plus 854.22 minus 181.66, for a total of 1,899.7 kcal/day, roughly 92 kcal or about 5 percent higher than Mifflin-St Jeor, consistent with the overestimation bias documented in the Frankenfield data.

Katch-McArdle first requires lean body mass: 85 times (1 minus 0.15) equals 72.25 kg. Then BMR equals 370 plus 21.6 times 72.25, which is 370 plus 1,560.6, for a total of 1,930.6 kcal/day, the highest of the three in this example, because this hypothetical subject is relatively lean and muscular for his weight, exactly the population where Katch-McArdle is expected to diverge upward from weight-based formulas.

The three formulas disagree by well over 120 kcal/day for the exact same person. That is a meaningful gap once it compounds into a calorie deficit or surplus target applied over weeks and months, and it is exactly why the choice of equation matters more than most people assume. You can compare your own numbers across formulas using the Scinergy macro calculator, which shows the underlying BMR estimate alongside the final macro targets rather than hiding the math.

DEXA vs BIA vs calipers: does your body-fat input even matter

Katch-McArdle's entire output hinges on one number, lean body mass, calculated as total weight times one minus body fat percent. Because body fat measurement methods vary enormously in precision, a Katch-McArdle estimate can only be as good as its input. DEXA, dual-energy X-ray absorptiometry, is generally considered the practical gold standard for body composition in non-research clinical settings, with typical error margins of around 1 to 3 percentage points of body fat. Bioelectrical impedance analysis, including consumer smart scales, is convenient but sensitive to hydration status, recent food or exercise, and device quality, with error margins that can exceed 5 to 8 percentage points, sometimes more in individuals with atypical body composition. Skinfold calipers depend heavily on technician skill and consistency of measurement site; in experienced hands they can approach DEXA-level accuracy for group averages, but individual error can still run several percentage points, and results are highly operator dependent.

A useful way to think about the stakes: an error of just 5 percentage points of body fat on an 85kg person shifts lean body mass by over 4kg, which shifts the Katch-McArdle BMR estimate by roughly 90 kcal, comparable in size to the entire accuracy gap between Mifflin-St Jeor and Harris-Benedict in the Frankenfield data. In other words, Katch-McArdle is only as trustworthy as its input, and a rough bioelectrical impedance scan can make it less accurate overall than simply using Mifflin-St Jeor.

The practical recommendation: use Mifflin-St Jeor as the default unless you have a reasonably recent DEXA scan, within the last 3 to 6 months, or a highly consistent caliper protocol from an experienced technician. A single bioelectrical impedance reading from a bathroom scale is not reliable enough to justify switching to Katch-McArdle over Mifflin-St Jeor, because the error introduced by a poor body-fat estimate can exceed the accuracy benefit the lean-mass equation is supposed to provide. Once you have a trustworthy BMR number from either formula, the next step is applying an activity multiplier to reach total daily energy expenditure, which carries its own accuracy problems covered in our activity multiplier deep dive. And if you are perimenopausal or postmenopausal, age-related lean mass loss changes both the BMR estimate and the deficit you can safely run, which we cover in our macros for perimenopause guide.