BMR vs RMR Formulas 2025: Mifflin-St Jeor vs Harris-Benedict Revised Equations

Basal and resting metabolic rate equations are among the most frequently cited formulas in nutritional epidemiology and human energy metabolism literature. Multiple competing prediction equations exist, spanning publication dates from 1918 through 1990, input variables ranging from simple weight-height-age to lean body mass. The academic literature contains dozens of formula variants, and the terminology around BMR versus RMR is often used inconsistently across sources. This reference page documents the three most widely used equations in current 2025 clinical and public health practice, explains the technical definitions of BMR versus RMR measurement conditions, presents the exact regression coefficients for Harris-Benedict original (1918), Harris-Benedict revised (1984 Roza-Shizgal), Mifflin-St Jeor (1990), and Katch-McArdle, provides three fully worked calculation examples with real numeric inputs, and compares the published accuracy of each formula against indirect calorimetry gold standard measurements including R² and RMSE values from the peer-reviewed validation literature.

Readers can compute BMR and RMR values using any of the three formulas described on this page in the BMR Calculator, which implements MSJ, Harris-Benedict revised, and Katch-McArdle side-by-side for comparison. Total Daily Energy Expenditure (TDEE) is calculated by multiplying BMR or RMR by an activity factor multiplier, as documented in the companion TDEE activity factor guide and implemented in the TDEE Calculator.

BMR vs RMR: What's the Difference (Definitions)

The terms BMR (Basal Metabolic Rate) and RMR (Resting Metabolic Rate) are frequently used interchangeably in popular writing, but they refer to measurements made under technically different laboratory conditions. The distinction between the two terms is important for understanding the published validation literature, even though many prediction equations are often applied to both labels in common clinical and wellness use.

BMR is measured under the most restrictive possible set of conditions. The standard BMR measurement protocol requires the subject to be in a thermoneutral environment (approximately 24-27°C, the temperature range where metabolic rate is minimized because the body does not work to generate or dissipate heat), to have fasted for 10 to 12 hours overnight, to have abstained from any vigorous physical activity for the preceding 12 to 24 hours, to have consumed no caffeine, nicotine, or other stimulant substances, to be lying motionless in a supine position, and to be either asleep or in the immediate post-awakening state before any voluntary muscle activity begins. BMR represents the minimum rate of energy expenditure required to sustain vital organ function under these strictly controlled conditions.

RMR is measured under less restrictive conditions. The typical RMR measurement protocol requires 3 to 4 hours of fasting (post-prandial but not overnight fasted), no vigorous exercise in the preceding few hours, and the subject resting quietly seated or supine in a comfortable environment. Because the thermic effect of food from recent meals and the less strict fasting requirement, measured RMR values are typically reported in the literature as being 5% to 15% higher than measured BMR values for the same individual. In common everyday usage and in many calculation tools, however, the terms BMR and RMR are used interchangeably to refer to the output of any of the prediction equations documented on this page. The FAO/WHO/UNU 2004 expert consultation report uses BMR as the umbrella term for the resting energy output category, regardless of strict measurement protocol.

Harris-Benedict 1918 Original Equations

The Harris-Benedict equations are the oldest continuously used BMR prediction formulas in modern nutrition science. They were published in two companion papers by James Arthur Harris and Francis Gano Benedict in 1918 (for males) and 1919 (for females) in the Proceedings of the National Academy of Sciences and the Carnegie Institution of Washington publication series. The original study sample consisted of indirect calorimetry measurements on 239 adult subjects (136 men and 103 women), primarily of European ancestry and residing in the Northeastern United States. The equations used multiple linear regression to predict 24-hour basal energy expenditure from measured body weight, height, and chronological age.

The original 1918 Harris-Benedict male equation for BMR in kilocalories per day is: BMR = 66.4730 + 13.7516 × weight(kg) + 5.0033 × height(cm) - 6.7550 × age(years). The original 1919 female equation is: BMR = 655.0955 + 9.5634 × weight(kg) + 1.8496 × height(cm) - 4.6756 × age(years). These coefficients—the 13.75 weight multiplier for males and the 9.56 weight multiplier for females—are the numbers most commonly associated with the label "Harris-Benedict original" in historical references. The equations were remarkably successful for their era and remained in widespread clinical use for more than six decades before the 1984 Roza-Shizgal revision documented measurable systematic bias when the 1918 coefficients were applied to populations outside the original 1918 sample demographic range.

1984 Revised Roza-Shizgal Update

In 1984, D. Roza and H. M. Shizgal published a re-estimation of the Harris-Benedict regression coefficients using a larger and more demographically diverse dataset of indirect calorimetry measurements. Their paper, published in the American Journal of Clinical Nutrition, demonstrated that the 1918 coefficients introduced measurable bias when applied to subjects from cohorts that differed demographically from the original 1918 study population. The 1984 revision adjusted all coefficients—intercept, weight, height, and age—for both sexes, producing the revised Harris-Benedict equations that are labeled "revised Harris-Benedict" or "Roza-Shizgal revised" in the modern literature.

The 1984 revised Roza-Shizgal male equation is: BMR = 88.362 + 13.397 × weight(kg) + 4.799 × height(cm) - 5.677 × age(years). Compare the 1984 female revised equation is: BMR = 447.593 + 9.247 × weight(kg) + 3.098 × height(cm) - 4.330 × age(years). The key numeric differences from 1918 vs 1984 are the male weight coefficient dropped from 13.7516 to 13.397, the male height coefficient dropped from 5.0033 to 4.799, and the female weight coefficient dropped from 9.5634 to 9.247. These coefficient changes reduced the systematic overprediction documented in certain demographic subgroups. The revised equations found that the 1918 original had tended to systematically overestimate BMR in later twentieth century subjects, likely due in part to secular changes in population body composition and the original narrow sample demographics.

Mifflin-St Jeor 1990: Today's De-facto Standard

The Mifflin-St Jeor (MSJ) equations were published in 1990 by Mark D. Mifflin, Sachiko T. St Jeor, Lisa A. Hill, Barbara J. Scott, Sandy A. Daugherty, and Yolanda O. Koh in the American Journal of Clinical Nutrition. The study used a sample of 498 indirect calorimetry measurements carefully selected to represent a broader range of age, weight, and body mass index values than prior datasets used in the 1918 and 1984 studies. The authors developed both simplified linear regression form deliberately to have fewer coefficient places, and simpler structure while maintaining or improving accuracy relative to the Harris-Benedict revised and competing equations.

The Mifflin-St Jeor equation published equations are as follows. Male MSJ male: BMR (kcal/day) = 10 × weight(kg) + 6.25 × height(cm) - 5 × age(years) + 5. MSJ female: BMR (kcal/day) = 10 × weight(kg) + 6.25 × height(cm) - 5 × age(years) - 161. The reader will observe that the weight, height, and age coefficients are identical between sexes; only the constant term differs (+5 for males, -161 for females), a difference of 166 kcal/day between the sex-specific intercepts. This symmetric structural simplicity is one of the practical advantages of the MSJ formulation. An evidence review conducted by the Academy of Nutrition and Dietetics (formerly the American Dietetic Association) reviewed all major BMR prediction equations and concluded that Mifflin-St Jeor was the most accurate prediction equation for use in clinical and public health settings, producing the lowest average error across populations typically tested. This Academy position, along with replication in multiple subsequent independent validations, has established Mifflin-St Jeor as the de facto standard BMR prediction equation as of 2025.

Katch-McArdle Formula for Lean-Mass Inputs

The Katch-McArdle resting metabolic rate formula differs structurally from both Mifflin-St Jeor and Harris-Benedict because it does not use weight, height, age, or sex as direct predictor variables. Instead, Katch-McArdle uses a single predictor: lean body mass (LBM) measured in kilograms. The formula was published by Frank I. Katch and Victor L. Katch in the exercise science and body composition literature. Its structural equation is RMR (kcal/day) = 370 + 21.6 × lean body mass(kg).

Because lean body mass is the primary predictor, using Katch-McArdle first requires an estimate of lean body mass. The most common method for obtaining this estimate is to subtract estimated fat mass from total body weight using a separately measured or estimated body fat percentage: LBM (kg) = weight (kg) × (1 - body_fat_fraction_decimal). For example, an individual with a measured or estimated body fat percentage of 20% has a body fat fraction of 0.20 and an LBM fraction of 0.80, so LBM = weight × 0.80. Because the accuracy of Katch-McArdle depends on the accuracy of the input LBM value; if body composition diverges from typical population proportions and the LBM estimate is inaccurate, then the resulting RMR estimate will carry that uncertainty forward. For individuals where a reasonably accurate LBM is available (e.g., from bioelectrical impedance, skinfold anthropometry, or other assessment), Katch-McArdle can differ materially from MSJ or Harris-Benedict estimates, particularly for individuals with very high or very low lean mass proportions.

Worked Examples: Three Cases

Example 1: 70 kg, 175 cm, 30-Year-Old Female

Compute the first worked example uses Mifflin-St Jeor, Harris-Benedict revised (Roza-Shizgal 1984), and Katch-McArdle for a 30-year-old female weighing 70 kilograms with a height of 175 centimeters. Mifflin-St Jeor female: (10 × 70) + (6.25 × 175) - (5 × 30) - 161 = 700 + 1093.75 - 150 - 161 = 1482.75 kcal/day. Harris-Benedict revised 1984 female: 447.593 + (9.247 × 70) + (3.098 × 175) - (4.330 × 30) = 447.593 + 647.29 + 542.15 - 129.9 = 1507.13 kcal/day. For Katch-McArdle, assume separately estimated 24% body fat: LBM = 70 × (1 - 0.24) = 70 × 0.76 = 53.2 kg. RMR = 370 + (21.6 × 53.2) = 370 + 1149.12 = 1519.12 kcal/day.

Example 2: 85 kg, 180 cm, 25-Year-Old Male

The second worked example uses a 25-year-old male, 85 kilograms, 180 centimeters. Mifflin-St Jeor male: (10 × 85) + (6.25 × 180) - (5 × 25) + 5 = 850 + 1125 - 125 + 5 = 1855 kcal/day. Harris-Benedict 1984 revised male: 88.362 + (13.397 × 85) + (4.799 × 180) - (5.677 × 25) = 88.362 + 1138.745 + 863.82 - 141.925 = 1949.00 kcal/day. Katch-McArdle with an estimated 18% body fat: LBM = 85 × 0.82 = 69.7 kg. RMR = 370 + (21.6 × 69.7) = 370 + 1505.52 = 1875.52 kcal/day.

Example 3: Body Fat Percentage → LBM → Katch-McArdle Full Worked Case

The third example demonstrates a full Katch-McArdle calculation from raw body fat percentage input without a body fat percentage measurement: a 40-year-old, 92 kg, 178 cm, 28% body fat. First compute LBM: 92 × (1 - 0.28) = 92 × 0.72 = 66.24 kg lean body mass. Katch-McArdle: 370 + (21.6 × 66.24) = 370 + 1430.78 = 1800.78 kcal/day. Now compare Mifflin-St Jeor female: 10 × 92 + 6.25 × 178 - 5 × 40 - 161 = 920 + 1112.5 - 200 - 161 = 1671.5 kcal/day. Here the Katch-McArdle estimate is approximately 129 kcal higher than the MSJ estimate, reflecting the information about body composition carried by the LBM input; MSJ infers composition only indirectly via total weight, height, age, and sex.

Formula Accuracy Comparison Table

The table below summarizes published validation study accuracy metrics reported in peer-reviewed BMR equation validation studies. R² is the proportion of variance in measured indirect calorimetry RMR values explained by each equation (higher = closer to 1 = better). RMSE is root mean squared error in kilocalories per day; lower values indicate smaller average prediction error. The ranges reported represent the approximate central tendency ranges reported across multiple studies in the literature; individual studies report individual ranges vary by study population characteristics.

Known Limitations: Equation Drift at BMI Extremes

All BMR and RMR prediction equations are population regression models. The coefficients were fit by minimizing squared prediction error across a training sample population. The error distribution around the regression line has tails. For individuals at the tails of the body mass index distribution, prediction error is documented to increase in published validation studies. Specifically, at BMI values below approximately 16 kg/m² and above approximately 40 kg/m², the correlation between predicted and measured R² with measured values decreases and the magnitude and direction of systematic bias can vary by study and by subgroup. The Academy of Nutrition and Dietetics evidence review notes that for individuals with BMI > 40, indirect calorimetry provides more accurate values when available. This limitation is inherent to the regression framework rather than any particular equation. Individual estimated output is not a substitute for measurement, any more than a population regression line is a substitute for an individually measured value.

Additional limitations include: (2) the input variables capture only some of the known factors that contribute to between-individual variation in resting energy expenditure. Genetic factors, thyroid axis function, body composition beyond lean vs fat ratio, recent weight history, medication use, and thermogenic effects of sympathetic nervous system tone all contribute to measured RMR variance and are not captured in weight-height-age-sex input variables. Published heritability studies document that a substantial fraction of between-individual RMR variance remains after accounting for age, sex, weight, and height. (3) the equations were trained primarily on populations of primarily European and US study participants in the original studies; replication studies in some non-Western ancestral groups have in some cohorts reported varying levels of systematic average differences, though MSJ generally performs the best available among existing equations. (4) they provide point estimates; any individual person's measured RMR can differ from the predicted value by more than the population RMSE.

Data Source Citations