How accurate are the formulas behind your calorie target?

Harris JA, Benedict FG. Proceedings of the National Academy of Sciences

Harris and Benedict derived the first multivariable basal-metabolism prediction equations from indirect calorimetry on 239 subjects (136 men, 108 women) using height, weight, age and sex — the template every later BMR equation refines.

Mifflin MD, St Jeor ST, Hill LA, Scott BJ, Daugherty SA, Koh YO. The American Journal of Clinical Nutrition

The original Mifflin-St Jeor derivation, built from indirect calorimetry on 498 healthy adults, produced a weight/height/age/sex equation explaining 71% of resting-energy variance (R-squared = 0.71) and improved on Harris-Benedict, which overestimated measured expenditure by about 5%.

Roza AM, Shizgal HM. The American Journal of Clinical Nutrition

Re-deriving Harris-Benedict from indirect calorimetry on 337 subjects, Roza and Shizgal reported it estimates a normal person’s resting energy expenditure with a precision of about 14%, and that expenditure tracks body cell mass independent of age and sex.

Frankenfield DC. Clinical Nutrition

Across 337 community-living adults the Mifflin-St Jeor equation was statistically unbiased (95% CI -26 to +8 kcal/day) and fell within 10% of measured RMR for 82% overall — but that accuracy dropped from 87% in non-obese to 75% in obese volunteers.

Pavlidou E, Papadopoulou SK, Seroglou K, Giaginis C. Metabolites

A 2023 re-derivation validated against indirect calorimetry reported its revised Harris-Benedict equation landed within 10% of measured RMR for 67.5% of men and 59.1% of women, with mean bias around 8-9% and a root-mean-square error near 176-205 kcal/day.

Pureza IRDOM, Macena ML, Silva Junior AE, Praxedes DRS, Florencio TMMT, Bueno NB. Archives of Endocrinology and Metabolism

In low-income obese women no equation met the ±10% standard: Mifflin-St Jeor (bias -4.1%) fell within 10% for only 33.8% of subjects, while Henry-Rees performed best (bias -0.8%, 42.3% within 10%) — a reminder that equation accuracy is population-specific.

Frankenfield D, Roth-Yousey L, Compher C. Journal of the American Dietetic Association

In this American Dietetic Association systematic review the Mifflin-St Jeor equation was the most reliable, landing within 10% of measured RMR in 82% of nonobese and 70% of obese adults (versus 69% and 64% for Harris-Benedict) and beating Owen and WHO/FAO/UNU, with the narrowest error range.

O’Neill JER, Corish CA, Horner K. Sports Medicine

In an athlete meta-analysis, five equations (Cunningham, Harris-Benedict, De Lorenzo, Ten-Haaf) showed no significant bias versus measured RMR; the body-weight-based Ten-Haaf equation was most precise, placing 80.2% of athletes within ±10% (the others 40.7-63.7%).

Jagim AR, Camic CL, Kisiolek J, Luedke J, Erickson J, Jones MT, Oliver JM. Journal of Strength and Conditioning Research

Among 50 athletes every equation underestimated measured RMR, but the fat-free-mass-based Cunningham equation carried the smallest mean error (-165 kcal/day) — evidence that knowing body composition helps most in lean, muscular people.

Tinsley GM, Graybeal AJ, Moore ML. Applied Physiology, Nutrition, and Metabolism

In 27 muscular physique athletes (men ~12.5% body fat, women ~19.2%) existing weight- and fat-free-mass-based equations generally underestimated measured RMR and lacked acceptable validity, so the authors had to derive new physique-athlete-specific equations.

Prado-Nóvoa O, Reina-Pérez I, Castets P, Kramer DK, et al.. Scientific Reports

Across 56 adults with total energy expenditure measured by doubly-labeled water, the best of 10 activity-factor equations still showed a 15.15% mean absolute percentage error and classified only 42.86% of people accurately, erring more for women and more active people.

Porter J, Nguo K, Collins J, Kellow N, Huggins CE, Gibson S, Truby H, et al.. The American Journal of Clinical Nutrition

In 1,488 doubly-labeled-water records from adults aged 65+, RMR equations multiplied by a measured activity level had small mean bias but very wide limits of agreement (about ±1,350 kJ/day) and systematically under-predicted high and over-predicted low expenders.

National Academies of Sciences, Engineering, and Medicine. National Academies Press

The 2023 Estimated Energy Requirement equations — derived from roughly 8,600 doubly-labeled-water observations with four activity-level categories — carry standard prediction errors of 96-342 kcal/day across age and sex strata, quantifying the residual error of any activity-factor target.

Westerterp KR. Frontiers in Physiology

Activity is the most variable part of daily expenditure: sustainable physical-activity level spans roughly 1.1 to 2.5, with population means near 1.70-1.77 — variability that a single fixed activity multiplier cannot capture for any one person.

Pontzer H, Yamada Y, Sagayama H, Ainslie PN, Speakman JR, et al. (IAEA DLW Database). Science

Analyzing 6,600+ people aged 8 days to 95 years, fat-free-mass-adjusted total energy expenditure stayed stable across adulthood (ages 20-60) and only declined later — so the common "metabolism crashes at 30" assumption built into many targets is unsupported.

Leibel RL, Rosenbaum M, Hirsch J. The New England Journal of Medicine

Maintaining a body weight 10% below baseline cut total energy expenditure by about 6 kcal/kg fat-free-mass/day in never-obese and 8 in obese subjects — beyond what the lost tissue alone predicts, the founding measurement of adaptive thermogenesis.

Rosenbaum M, Hirsch J, Gallagher DA, Leibel RL. The American Journal of Clinical Nutrition

In people maintaining a ≥10% weight loss for over a year, total and non-resting energy expenditure stayed significantly below the values predicted from body composition — adaptive thermogenesis persists well beyond active weight loss, not just during it.

Rosenbaum M, Leibel RL. International Journal of Obesity

This review reports that maintaining a ≥10% weight loss runs energy expenditure roughly 300-400 kcal/day below what fat and lean mass predict — so a reduced-weight person needs meaningfully fewer calories than a never-obese person of identical size.

Fothergill E, Guo J, Howard L, Kerns JC, Knuth ND, Brychta R, Hall KD, et al.. Obesity

In 14 "Biggest Loser" contestants resting metabolic rate fell 610 kcal/day after the 30-week competition; six years later, despite regaining most of the lost weight, it was still about 499 kcal/day below the rate predicted for their body size.

Nunes CL, Casanova N, Francisco R, Bosy-Westphal A, Hopkins M, Sardinha LB, Silva AM. British Journal of Nutrition

Across 33 studies (2,528 participants), 23 of 29 that measured resting expenditure found significant adaptive thermogenesis after weight loss — but the effect was often small or non-significant in the higher-quality designs, so it is real yet methodology-sensitive in size.

Hall KD, Sacks G, Chandramohan D, Chow CC, Wang YC, Gortmaker SL, Swinburn BA. The Lancet

Hall’s validated dynamic energy-balance model (the basis of the NIH Body Weight Planner) shows the static "3,500 kcal per pound" rule overpredicts weight loss by roughly 100% over a year, because expenditure falls as weight drops; a lasting 100 kJ/day change yields about 1 kg, half realized in ~1 year.

Thomas DM, Martin CK, Heymsfield S, Redman LM, Schoeller DA, Levine JA. International Journal of Obesity

Across seven supervised-feeding studies (103 adults, ~65 days) the 3,500-kcal rule predicted 27.6 lb of loss versus 20.1 lb actually lost — a 7.4 lb overestimate — while a dynamic model predicted 18.4 vs 17.0 lb observed.

Hall KD, Chow CC. International Journal of Obesity

Hall and Chow argue the rule’s deepest flaw is treating energy balance as static: by ignoring the metabolic adaptation that progressively shrinks the deficit, it predicts indefinite linear loss with no plateau and overstates real weight change.

Hall KD. International Journal of Obesity

Using a Forbes-based model, Hall shows the cumulative deficit per unit weight loss is not a fixed 3,500 kcal/lb but rises with starting body fat — the rule roughly fits very obese people yet overestimates the deficit needed in leaner ones.

Westerterp KR. European Journal of Applied Physiology

Validated against the respiration chamber, the doubly-labeled-water method is accurate to within about 2% with a precision of 4-8% — the gold standard for free-living energy expenditure still carries several percent of its own error.

Henriksen HB, Knudsen KEB, et al.. Physiological Reports

Repeated whole-room indirect calorimetry of resting metabolic rate measured 24 hours apart showed a between-day coefficient of variation of 2.68% — so even a directly measured baseline shifts a few percent from one day to the next.

Donahoo WT, Levine JA, Melanson EL. Current Opinion in Clinical Nutrition and Metabolic Care

This review puts the coefficient of variation for resting metabolic rate at roughly 5-8% (and 5-10% for 24-hour expenditure by room calorimeter), so even a directly measured calorie baseline reproducibly varies several percent between measurements.

Johnstone AM, Murison SD, Duncan JS, Rance KA, Speakman JR. The American Journal of Clinical Nutrition

In 150 adults, fat-free mass explained 63% of the between-person variation in basal metabolic rate, fat mass 6% and age 2% — leaving about 26% unexplained, which is why two people of identical size and composition can have materially different true metabolic rates.