SYNTH-87 · Guide

Calories on the label vs what the body gets

A label or app number is an Atwater estimate, a workable starting point for a mixed diet (practice, not measured) and wrong in a known direction for some foods (whole nuts, very fibre-rich whole-food diets); on an ordinary diet people differ little in how much of a food they absorb and a lot in what their body does with a surplus; track as a guide, judge by the body-mass trend, performance and recovery, and do not chase precision

B-limited, mixed evidenceMostly from non-athletes and recreational people

The decision this framework makes

How much weight to put on a calorie number from a label, a database or an app; when it misleads and in which direction; and what an athlete should steer by instead — the body-mass trend, performance and recovery — so that intake is neither under-fuelled by false precision nor left to guesswork.

Key Findings — what a calorie number is and how far it can be off

1

A label calorie is a calculation, not a burned-and-measured value. Labels and databases apply the Atwater factors — the general 4, 9 and 4 kcal/g for protein, fat and carbohydrate, or food-specific factors — which assume an average digestibility for each class of food (REF-1124, B+, full text). The result is an estimate of the energy the body can use, not a measurement of it.

2

For whole nuts the estimate is too high. Whole almonds: 129 kcal measured vs 168-170 kcal predicted per 28-g serving; the Atwater factors overestimated by 32% (REF-1124, B+, n = 18). Walnuts: 146 vs 185 kcal per serving, 21% less than predicted (REF-1125, B+, n = 18, abstract). The measured reason for almonds is fat that stays in the stool (faecal fat 1.7 → 6.3 → 10.8 g/d with 0, 42 and 84 g/d).

3

Whole, fibre-rich diets are over-counted; highly digestible diets less so. A diet designed for the gut microbes (fibre, resistant starch, whole nuts instead of nut butter, few processed foods) lost an extra 116 ± 56 kcal/day in faeces: 89.5% of intake metabolizable vs 95.4% on a Western diet the standard equations rated equal (REF-1131, B+, n = 17).

The authors of REF-1130 (B+) calculate, without measuring it, that their unprocessed diet at ~46 g/d of fibre was over-counted by ~330 kcal/d.

4

Labels can also understate. Hospital meals analysed by bomb calorimetry carried more energy than their product labels stated (REF-1126, B; the size is shown only in a figure).

5

Processing changed how much people ate. With menus matched for calories, macronutrients, sugar, sodium and fibre, adults eating freely took in 508 ± 106 kcal/day more on ultra-processed food and gained 0.9 kg in 2 weeks, against a 0.9-kg loss on unprocessed food (REF-1130, B+, n = 20) (absorption was not measured; if the authors' ~330 kcal/d estimate is right, the real gap in energy absorbed was larger, not smaller).

A whole-food sandwich cost 137 vs 73.1 kcal to digest (19.9% vs 10.7% of meal energy) in one single-meal study (REF-1129, B), but 24-h chamber expenditure did not show that gap (REF-1130).

6

People absorb about the same share of the same food. Fed identical meals, lean and obese men lost 4.9 ± 1.8% vs 4.8 ± 1.4% of intake in stool at 2400 kcal/d; individuals ranged 2.1-9.2% (REF-1126, B, n = 21) — arithmetic: roughly 50-220 kcal/d at that intake. The ~150 kcal microbiota link was an association in the lean men only. On a Western diet people differed little (94.1-97.0% metabolizable); on the whole-food diet much more (84.2-96.1%) (REF-1131, B+).

7

People differ a lot in what a surplus does to them. The identical 1000 kcal/day surplus for 84 days (84,000 kcal) produced 4.3 to 13.3 kg of gain (mean 8.1 kg), with about three times more variance among twin pairs than within them (REF-1127, B, n = 24, abstract).

In 16 nonobese adults overfed 1000 kcal/day for 8 weeks, two-thirds of the rise in expenditure was non-exercise activity (NEAT), fat storage differed 10-fold, and the NEAT change predicted resistance to fat gain (r = 0.77) (REF-1128, B, abstract).

So, by inference across studies: on an ordinary diet, absorption differences between people are a few per cent of intake (wider, up to ~12 points, on a very high-fibre whole-food diet). Differences in partitioning and spontaneous activity can triple the result of the same surplus. The second is why two athletes eating the "same" calories end up in different places, and why the athlete's own trend beats any formula.

Default (what to prescribe first)

1

Use the log as a guide, not a measurement. Any reputable app or label is good enough to set a starting intake (the base's practice; no entry here measures label error for a whole mixed diet); its errors run mostly toward over-counting what the body absorbs for whole nuts and fibre-rich whole foods (REF-1124, REF-1125, REF-1131, B+).

2

Judge the intake by outcomes. Body-mass trend reviewed every 2-3 weeks, key-session numbers and recovery — the review cadence is the house default carried by SYNTH-54 (B), not a finding. If the trend disagrees with the log, change the intake by the trend; do not re-audit the log to the gram.

3

Expect your own response, not the average. The same surplus lands differently from person to person, mainly after absorption (REF-1127, REF-1128, B; a deficit was not studied, so the same for a deficit is inference); a friend's numbers or a formula's prediction are a starting point only.

4

Fuel training first. In a hard block, carbohydrate and protein targets come from SYNTH-69 (B), and falling body mass with worse sessions or recovery means eat more, whatever the log says.

Decision (situation → what to do)

SituationWhat to doSource and how sure
"Is my app's calorie number accurate?"An estimate, good enough to start from (practice, not measured for mixed diets); over-counts whole nuts (Atwater overestimate of 32% for almonds; walnuts measured 21% below prediction) and very fibre-rich whole-food diets (89.5% vs 95.4% of intake metabolizable in one designed diet); a label can also understateREF-1124 B+, REF-1125 B+ (abstract), REF-1131 B+, REF-1126 B; all general population — transfer
"Does the same food give different people different energy?"A little in the gut (stool loss 4.9% vs 4.8% of intake, lean vs obese; individuals 2.1-9.2%; more on a very high-fibre diet, 84.2–96.1% metabolizable, REF-1131); a lot after absorption (4.3-13.3 kg from one identical surplus)REF-1126 B; REF-1131 B+; REF-1127 B, REF-1128 B (abstracts)
Cutting weightScreen for low energy availability first (SYNTH-54); then set intake from the log and steer by the 2-3-week body-mass trend; a mostly unprocessed diet may make intake easier to control (transfer); do not correct labels by handSYNTH-54 B (cadence = house default); REF-341 A (REDs screen); REF-1130 B+ (+508 kcal/day ad libitum on ultra-processed food)
"In a deficit on paper but not losing"Trust the trend over the log; a whole-food diet means the log, if anything, over-counts, so look at portion logging and at reduced moving outside training before cutting harder; screen for low energy availability before any further deficitREF-1131 B+; REF-1128 B (NEAT, inference for the deficit direction, not tested); SYNTH-54 B and the REDs screen in REF-341 A
Hard block, camp, multi-day eventDo not cap intake at a calorie target; fibre-rich whole food is generously counted, so a log that "covers" training can leave a shortfall; energy-dense, lower-fibre carbohydrate makes high targets reachable (inference); in very long multi-day efforts expenditure can outrun what the gut supplies (a group-level ceiling near 2.5 × BMR, not an individual limit; SYNTH-24)REF-1131 B+; SYNTH-69 B; SYNTH-24 B (REF-958 B)
Gaining mass in the off-seasonSmall surplus, steer by trend; large surpluses mostly add fatREF-1088 B, REF-1089 B+; spread of response REF-1127 B
"Whole food burns more calories to digest"Do not count on it: one single-meal study vs no 24-h difference in a chamberREF-1129 B vs REF-1130 B+
"My microbiome steals or adds calories"The measured effects average about 100–150 kcal a day (116 ± 56 kcal/day extra faecal loss on a designed whole-food diet; ≈150 kcal per 20% Firmicutes shift, an association in lean men only) and come from what is eaten; no test or probiotic in the base changes thisREF-1126 B (association, lean only); REF-1131 B+ (designed diet, causes not separated)

Practical Application

"Are calories a spherical horse in a vacuum?" No — but a label calorie is an average estimate, not a measurement.For a mixed diet it is a workable starting point (practice, not measured); for whole almonds the Atwater factors overestimated by 32% (129 vs 168-170 kcal per 28 g; REF-1124, B+), for walnuts measured energy was 21% below prediction (REF-1125, B+), and a whole-food, fibre-rich diet lost 116 ± 56 kcal/day more in faeces than a Western diet counted as equal (REF-1131, B+).The same food gives different people only modestly different energy (stool loss 4.9% vs 4.8% of intake, lean vs obese; REF-1126, B); what differs a lot is what the body does with a surplus — 4.3 to 13.3 kg from one identical 84,000-kcal surplus (REF-1127, B). So track as a guide and let the body-mass trend, performance and recovery decide.
Pick the steering signals: body mass on the same scale under the same conditions, read as a trend over 2-3 weeks (house default via SYNTH-54); key-session power or pace; sleep, mood and soreness. A food diary is a poor instrument for energy availability: in 59 elite distance athletes, reproductive function was a better marker than diary-based energy availability (REF-343, B).
Where the label errs, it errs mostly in the athlete's favour for a cut and against them in a heavy block: whole nuts and fibre-rich whole food are over-counted (REF-1124, REF-1125, REF-1131); the same over-count can hide an under-fuelled training week (SYNTH-69).
Whole food is a tool for appetite, not for maths: free-feeding adults ate 508 ± 106 kcal/day more on ultra-processed food (REF-1130, B+); use whole food when the aim is to eat less, and accept energy-dense carbohydrate when the aim is to eat enough (inference).
If tracking itself starts to cost more than it gives, stop counting: a systematic review found 27 of 65 articles examining disordered eating linked to health and fitness apps and trackers, with several null studies (REF-405, B+, general consumers); stopping is the base's own practice, not a finding.

What NOT to do

Do not weigh everything to the gram or re-calculate labels for nuts: the almond and walnut errors are tens of kilocalories per serving (39 kcal per walnut serving; REF-1125), well inside what the body-mass trend corrects.
Do not add a "thermic bonus" for whole food (REF-1129 B, not reproduced over 24 h in REF-1130 B+).
Do not buy microbiome tests or probiotics to change calorie harvest: no study in the base manipulated the microbes alone; REF-1126 is an association in lean men, and REF-1131 changed the whole diet and could not separate diet from microbes.
Do not give or follow a calorie target without the screen: low-energy-availability indicators end the weight conversation (SYNTH-54 B, REF-341 A).
Do not blame "a slow metabolism" or "bad absorption" before the trend is read: absorption differs little between people (REF-1126, B); differences in spontaneous activity and partitioning are larger (REF-1128, REF-1127, B) and show up in the trend, which is what to act on.
Do not trust a planned high-fibre whole-food day to cover a big training load on paper (REF-1131, B+; SYNTH-69).

Limitations

All eight studies are in untrained or unreported general-population adults (n = 16 to 24 each); none measured athletes, high training loads or race-day fuelling.
The two overfeeding studies behind the person-to-person claim (REF-1127, REF-1128) have no control group, 24 men and 16 volunteers, and are read from the abstract, as is REF-1125.
Only two nuts have measured energy here; other foods' label errors are not covered by the base.
The ~330 kcal/d over-count of a high-fibre diet in REF-1130 is the authors' calculation, not a measurement.
Nothing here tests how accurate app or watch estimates of exercise expenditure are, or NEAT during a deficit; those statements are inference.
The 2-3-week review cadence and "steer by the trend" are coaching practice, not tested rules.
How this entry was checked and changed
Why this level

(the food-level errors rest on four B+ controlled crossover feeding trials of 17-20 untrained adults; the person-to-person differences on three B studies, one correlational and two uncontrolled overfeeding studies, of 16 volunteers and of 24 men; no study was done in athletes, so every row is a transfer; the decision rules — track as a guide, judge by the trend, the 2-3-week review — are the base's practice)