TDEE Formulas Compared: Mifflin-St Jeor vs Harris-Benedict vs WHO/FAO vs Katch-McArdle

Dated caveat. As of 2026, the Frankenfield et al. 2005 comparison^[3] remains the definitive head-to-head of these formulas against indirect calorimetry. Subsequent comparisons have refined the error bands but not changed the ordering — Mifflin-St Jeor remains the most accurate at the population level.

What TDEE actually is

Total Daily Energy Expenditure is the sum of four components:

• BMR / RMR — energy your body uses at rest, roughly 60–75% of TDEE for most adults.
• TEF — thermic effect of food, roughly 10% of TDEE.
• NEAT — non-exercise activity thermogenesis (fidgeting, walking, chores), highly variable.
• EAT — exercise activity thermogenesis, exercise you deliberately do.

TDEE formulas calculate BMR directly, then multiply by an activity factor to account for the other three components. The activity factor is where most estimation error creeps in, not the BMR step.

Mifflin-St Jeor (1990)

Men:   BMR = 10 × weight_kg + 6.25 × height_cm − 5 × age + 5
Women: BMR = 10 × weight_kg + 6.25 × height_cm − 5 × age − 161

Fitted against indirect calorimetry in 498 healthy adults in 1990^[1]. Mifflin-St Jeor replaced Harris-Benedict as the ADA's preferred equation in clinical guidance because:

• The 1990 cohort was closer to modern adult body composition than Harris-Benedict's 1918 cohort.
• Mean absolute error against calorimetry was lower than any other formula in Frankenfield's 2005 review^[3].
• Within the study population, 82% of predictions fell within 10% of measured BMR.

This is the formula the TDEE Calculator and BMR Calculator default to.

Harris-Benedict (1918, revised 1984)

Original (1918):
  Men:   BMR = 66.5 + 13.75 × weight_kg + 5.003 × height_cm − 6.775 × age
  Women: BMR = 655.1 + 9.563 × weight_kg + 1.85 × height_cm − 4.676 × age

Revised Harris-Benedict (Roza & Shizgal 1984):
  Men:   BMR = 88.362 + 13.397 × weight_kg + 4.799 × height_cm − 5.677 × age
  Women: BMR = 447.593 + 9.247 × weight_kg + 3.098 × height_cm − 4.330 × age

Derived from a small 1918 cohort of primarily young Caucasian adults^[2]. The 1984 revision improved the fit marginally but did not address the core population bias. Frankenfield et al.^[3] found Harris-Benedict systematically overestimates BMR by about 5% in contemporary adults, particularly in women and older adults.

Still widely quoted because it's the oldest and most cited formula. Not the formula you should use unless you have a specific reason to match a historical reference point.

WHO/FAO/UNU (1985)

The World Health Organization commissioned a large cross-population review that produced age-bracketed formulas^[4]. For men aged 30–60:

Men 30–60:   BMR = 11.6 × weight_kg + 879
Women 30–60: BMR =  8.7 × weight_kg + 829

Simpler than Mifflin-St Jeor but less accurate for most individuals because it doesn't use height and averages over wider age brackets. Useful in population-level dietary planning, not in personal tracking.

Katch-McArdle

BMR = 370 + 21.6 × LBM_kg

The only formula in common use that scales on lean body mass rather than total weight^[5]. For athletic individuals with known body composition, Katch-McArdle outperforms Mifflin-St Jeor because fat-free mass is the actual driver of BMR — adipose tissue metabolises at a very different rate than muscle.

Requirement: you need a decent estimate of your lean body mass. If your LBM estimate is noisy (e.g. a bathroom BIA scale), Katch-McArdle's error compounds and it becomes less accurate than Mifflin-St Jeor. The Lean Body Mass Calculator gives you a defensible LBM estimate from bodyweight and body-fat percentage.

Head-to-head comparison

For a 35-year-old, 180 cm, 82 kg male at 15% body fat (LBM = 69.7 kg):

Formula                BMR (kcal)    TDEE (× 1.55 activity)    Δ vs Mifflin
──────────────────────────────────────────────────────────────────────────
Mifflin-St Jeor            1,770          2,744                 baseline
Harris-Benedict revised    1,833          2,841                 +3.5%
WHO/FAO/UNU (30–60 M)      1,830          2,837                 +3.4%
Katch-McArdle (LBM 69.7)   1,876          2,908                 +6.0%

For this lifter, the range is 2,744 to 2,908 kcal — about 160 kcal of spread, or a small banana's worth. Over a 6-week cut at a 500-kcal deficit target, that's the difference between losing 3.5 kg and losing 4.6 kg. Not trivial, not dramatic. Pick one, stick with it, and let the scale trend validate.

The activity multiplier is where real error lives

Activity multipliers typically look like this:

Sedentary       × 1.20    Desk job, no exercise
Lightly active  × 1.375   Light exercise 1–3 days/wk
Moderately      × 1.55    Moderate exercise 3–5 days/wk
Very active     × 1.725   Heavy exercise 6–7 days/wk
Extra active    × 1.90    Physical job + heavy training

These are population-averaged and can easily be off by 200–400 kcal in either direction for a specific person. Two people at the same BMR and the same labelled activity level can have very different NEAT and produce 15% divergent TDEEs. This is why “TDEE calculator said I eat 2,600 and I'm not losing at 2,100” happens regularly — the formula is giving you a starting point, not a ground truth.

Special cases the formulas handle poorly

Older adults (50+)

All four formulas under-predict BMR in adults over 50 to varying degrees. The lean-mass loss component is captured in Katch-McArdle (which scales on LBM directly) but underweighted in Mifflin-St Jeor. For older athletes with maintained muscle mass, Katch-McArdle is the defensible default; for older sedentary adults, Mifflin-St Jeor is fine.

Obese adults (BMI > 35)

Mifflin-St Jeor's fit degrades at the extremes of bodyweight. For individuals above BMI 35, using adjusted body weight (ABW = ideal bodyweight + 25% of excess) in the formula produces more defensible results than using actual bodyweight. Clinical dietitians often apply this correction routinely.

Very lean lifters (BF < 10% men, < 16% women)

Mifflin predictably undershoots for lean lifters because it uses total bodyweight — which includes fat mass that burns at ~4 kcal/kg/day — without upweighting the more metabolically active lean mass. For contest-lean athletes, Katch-McArdle is substantially more accurate.

Pregnancy and lactation

None of the standard formulas incorporate pregnancy or lactation energy costs. Add roughly 300 kcal/day to the computed TDEE during the second trimester, 450 kcal/day in the third, and 500 kcal/day during exclusive breastfeeding. These are WHO guidance and are conservative.

Adjusting for sleep and circadian state

BMR drifts with sleep debt and circadian state. Acute short-sleep nights don't change BMR meaningfully, but sustained patterns of 5-hour sleep can drop resting metabolic rate by 2–4% beyond what formulas predict. Another reason to treat TDEE as an estimate validated by bodyweight outcomes rather than a fixed number.

Validating your TDEE against reality

The honest procedure is:

1. Compute a baseline TDEE with Mifflin-St Jeor + the activity multiplier that best matches your routine.
2. Eat at that target for 10–14 days, tracking food honestly and weighing yourself daily (average the week).
3. If weight is stable, that's your actual TDEE. If it's drifting, adjust by the calorie equivalent: 1 kg change per week ≈ 1,000 kcal/day offset.

Metabolic adaptation matters here. Under sustained calorie restriction, BMR falls by a few percent beyond what the formulas predict^[6], so your measured TDEE at week 8 of a cut is lower than your measured TDEE at week 1. This is normal, not broken metabolism.

Which formula should you use?

• Default: Mifflin-St Jeor. Best population-level accuracy.
• Athletes with known lean mass: Katch-McArdle. Scales on the right variable.
• Quick-and-rough estimate without height: WHO/FAO/UNU. Good for back-of-envelope.
• Harris-Benedict: historical reference only.

Tools: TDEE Calculator, BMR Calculator, Macro Calculator.

Population boundaries of the source equations

Every TDEE formula in common use was fit on a specific cohort. Honest application means knowing whether you are in that cohort or extrapolating outside it:

• Mifflin-St Jeor (1990) — sample. 498 healthy adults, ages 19–78, 55% female. Recruited from a Nevada university city^[1]. The cohort was predominantly white, Hispanic, and Black American; Asian populations were under-represented. Multiple follow-up studies on East Asian and South Asian cohorts have shown Mifflin systematically over-predicts BMR by 3–7% in these groups. If you are of Asian descent, expect your true BMR to sit 50–150 kcal below the Mifflin output.
• Harris-Benedict (1918) — sample. 239 subjects recruited at Carnegie Nutrition Laboratory. Young adult Caucasian males and females, very different modern body composition (higher muscle mass relative to equivalent BMI today in most 2020s populations because of sedentary lifestyle shift). The 5% population-wide over-prediction documented by Frankenfield et al.^[3] reflects exactly this population drift.
• Katch-McArdle — sample. Derived from measurements on lean, trained adults. The 370 + 21.6×LBM linear fit^[5] holds at typical adult LBM ranges (40–90 kg) but extrapolates unreliably at extremes. For very lean contest-prep athletes (LBM > 100 kg) or very small adults (LBM < 35 kg), the linearity assumption starts to fail.
• WHO/FAO/UNU — sample. Aggregated across multiple regional cohorts but with large age brackets and no height term^[4]. Designed for population-level dietary planning, not individual prescription; the error band is wider than any of the other formulas for a given person.
• None of the formulas capture pregnancy, lactation, childhood/adolescence, clinical illness, or drug-induced metabolic shifts. For any of these populations, specialist guidance replaces all four.

Alternative-view framing: measured vs predicted

The underlying question the formulas answer is "what is your resting metabolic rate?" — which can also be measured directly. The measurement hierarchy from most to least expensive:

• Indirect calorimetry (gold standard). Fasted metabolic testing measures O₂ consumption and CO₂ production over 20–40 minutes. Cost: $150–350. Accuracy: ±3%. Available at most university exercise-science labs and increasingly at specialty clinics.
• Doubly-labelled water (research standard). 7–14 day free-living TDEE measurement via isotope tracking. Cost: $800–1,500 (rarely available clinically). Accuracy: ±2% for total TDEE, not just BMR.
• Wearable resting metabolism estimates. Consumer wearables' "resting metabolic rate" outputs are model-derived, typically Mifflin-style with wearable-specific activity corrections. They are not measurements and carry full formula error — despite the "your personal RMR" branding.
• Formula + bodyweight validation (what this article recommends). Zero cost, 2-week turnaround, ±5–10% accuracy for most people. The honest ceiling without going to a lab.

Worked example: cross-checking Mifflin against real-world bodyweight trend

A 45-year-old female, 168 cm, 65 kg, moderately active (3× strength + 2× Zone 2 weekly). Compute TDEE across the four formulas, then validate against 14 days of bodyweight-stable maintenance eating.

Formula               BMR    × 1.55 TDEE   Claimed maintenance
──────────────────────────────────────────────────────────────
Mifflin-St Jeor       1,325    2,054        2,050 kcal/day
Harris-Benedict rev.  1,400    2,170        2,170 kcal/day
WHO/FAO/UNU (30–60F)  1,390    2,155        2,155 kcal/day
Katch-McArdle@20%BF   1,494    2,316        2,316 kcal/day

14-day maintenance test eating 2,050 kcal (Mifflin default)
  Bodyweight day 1:   65.1 kg
  Bodyweight day 14:  65.8 kg  (+0.7 kg)
  → Implied actual TDEE ≈ 2,050 + (700g × 7700 kcal/kg ÷ 14) = 2,435 kcal

Calibrated TDEE: ~2,435 kcal, closest to Katch-McArdle estimate
  Likely explanation: NEAT higher than "moderate" multiplier assumed

This is a common pattern: formulas centered on sedentary-to-lightly-active populations under-estimate TDEE for recreationally active adults whose NEAT component is above average. The 400-kcal gap between the Mifflin default and real TDEE would produce a silent surplus on "maintenance," explaining the common frustration that a Mifflin-calibrated bulk fails to produce weight gain while the cut dialed in against the same number works fine. Validate, then trust the validated number.

Common failure modes

• Picking the activity multiplier aspirationally. "I go to the gym, I must be moderately active." A desk worker with 3× weekly strength sessions is typically 1.4× base — the exercise adds ~300 kcal/day, not the ~500 kcal the "moderate" 1.55 multiplier assumes. Mis-classify upward, your TDEE reads 150–250 kcal high, and the cut stalls.
• Treating TDEE as a fixed constant. Under sustained caloric restriction, BMR falls 5–10% beyond formula prediction through metabolic adaptation^[6]. The TDEE that matched reality at week 1 of a cut under-predicts by 100–200 kcal at week 10. Re-validate at the stall.
• Using total bodyweight in Katch-McArdle. The formula needs lean body mass, not bodyweight. Plugging 85 kg instead of the 70 kg LBM equivalent adds 320 kcal of fictional BMR. If LBM is noisy, use Mifflin instead.
• Switching formulas mid-cut. Starting with Mifflin, switching to Katch-McArdle when the scale stalls introduces a ~160 kcal target change that will reliably confuse the diagnosis of what's actually happening with the cut. Pick one formula, stick with it, adjust the number based on bodyweight trend — not based on switching formulas.
• Ignoring population-specific adjustments. For Asian-ancestry adults, subtracting 50–100 kcal from the Mifflin output before starting validation saves a predictable 2–3 weeks of "why isn't this working." The formula's population bias is documented; working around it is cheap.
• Trusting wearable-reported TDEE. Wearables combine a Mifflin-like BMR estimate with step-and-HR-based activity calorie estimates. The activity side routinely over-reports by 20–40% for steady-state cardio and under-reports resistance training. A wearable TDEE of 2,850 kcal for a recreational runner often matches real TDEE of 2,400–2,500. Validate against bodyweight, not against the wearable's display.
• Conflating TDEE with "calories to eat". TDEE is the total energy output; the eating target is TDEE + goal modifier (surplus, deficit, or maintenance). Athletes who hear "your TDEE is 2,800" and eat 2,800 expecting to lose weight are making a conceptual error — that's maintenance, not a cut.

Activity-multiplier calibration shortcuts

A pragmatic refinement to the standard 1.2 / 1.375 / 1.55 / 1.725 / 1.9 ladder: compute BMR from Mifflin, then add exercise calories additively instead of multiplicatively. This produces more defensible numbers for athletes with variable or spiky training weeks. Reference calorie costs for common work:

Activity (60 min)                    kcal range (70 kg adult)
─────────────────────────────────────────────────────────────
Resistance training (hypertrophy)    280–380
Resistance training (strength)       240–320
Moderate run (6:00/km pace)          580–680
Easy run (7:00/km pace)              440–520
Zone 2 cycling (steady)              460–580
Hard interval running (with W/CD)    520–640 + EPOC 80–140
Swimming (moderate)                  420–540
Yoga (steady)                        180–260
Walking (brisk, 5–6 km/h)            240–320

For a 70 kg adult with Mifflin BMR ~1,650 kcal, running 4× weekly (avg 600 kcal/session) and lifting 3× weekly (avg 300 kcal/session), weekly exercise = 3,300 kcal / 7 = 471 kcal/day. Non-exercise baseline: 1,650 × 1.3 = 2,145. TDEE: 2,145 + 471 = 2,616. Validate against bodyweight trend and adjust. This additive method typically lands within ±5% of measured TDEE; the multiplicative ladder can miss by 15%+ for spiky training schedules.